Development of S/MAR minicircles for enhanced and persistent transgene expression in the mouse liver
- First Online:
- Cite this article as:
- Argyros, O., Wong, S.P., Fedonidis, C. et al. J Mol Med (2011) 89: 515. doi:10.1007/s00109-010-0713-3
- 558 Views
We have previously described the development of a scaffold/matrix attachment region (S/MAR) episomal vector system for in vivo application and demonstrated its utility to sustain transgene expression in the mouse liver for at least 6 months following a single administration. Subsequently, we observed that transgene expression is sustained for the lifetime of the animal. The level of expression, however, does drop appreciably over time. We hypothesised that by eliminating the bacterial components in our vectors, we could improve their performance since bacterial sequences have been shown to be responsible for the immunotoxicity of the vector and the silencing of its expression when applied in vivo. We describe here the development of a minimally sized S/MAR vector, which is devoid of extraneous bacterial sequences. This minicircle vector comprises an expression cassette and an S/MAR moiety, providing higher and more sustained transgene expression for several months in the absence of selection, both in vitro and in vivo. In contrast to the expression of our original S/MAR plasmid vector, the novel S/MAR minicircle vectors mediate increased transgene expression, which becomes sustained at about twice the levels observed immediately after administration. These promising results demonstrate the utility of minimally sized S/MAR vectors for persistent, atoxic gene expression.
KeywordsScaffold/matrix attachment region (S/MAR)MinicirclePlasmidNon-viralGene therapyLiverHydrodynamic delivery
Improving the level and the duration of gene expression from DNA-based constructs is crucial to further progress in non-viral gene therapy. Typically, plasmid-mediated transgene expression reaches its maximum 12–48 h after delivery to the liver or the lungs but subsequently drops rapidly lasting rarely beyond 1 week, even though vector DNA is retained in cells of these tissues [1, 2]. Several factors have been suggested to explain the observed decline of gene expression. These include promoter CpG methylation , DNA heterochromatinisation  and inflammatory responses to immunostimulatory elements of pDNA involving cytokines TNFα, IFNγ and IL12 through the action of Toll-like receptor 9 . These cytokines can downregulate gene expression at a post-transcription level .
A recent study on plasmid DNA delivered to the mouse liver indicated that CpG motifs in the bacterial backbone contribute only little to transgene silencing and innate immune responses and that this type of silencing is only relevant to CpG-rich pDNA complexed with lipid nanoparticles . Another study by Riu et al. showed that both persistence and silencing of transgene expression in vivo were associated with specific histone modifications. The epigenetic events that led to expression silencing of an episomal plasmid vector were an increase in heterochromatin-associated histone modifications and a decrease in modifications typically associated with euchromatin. In contrast, a supercoiled vector devoid of a bacterial plasmid backbone (called minicircle) exhibited persistent expression and had a pattern of histone modifications consistent with euchromatin . Co-injection of both a circular bacterial backbone and a circular expression cassette showed that expression was not inhibited in trans, whereas a plasmid vector was silenced in cis by the linkage of an expression cassette to bacterial sequences. These data confirmed previous work showing that covalent linkage of the mammalian expression cassette to the bacterial backbone is of crucial importance for transgene silencing .
We reported that in vivo methylation of the cytomegalovirus and the liver-specific α1-antitrypsin (AAT) promoter results in transgene expression (luciferase) silencing in the murine liver, following hydrodynamic delivery of a plasmid vector . It is possible that methylation of the CpG motifs of the promoter could be an initial or an intermediate step, which directly affects gene expression by blocking access of transcription factors, or indirectly by promoting heterochromatin-associated histone modifications. In support of this theory are experiments showing that de novo DNA methylation can spread from an integrated retrovirus to flanking sequences in the mouse genome . In addition, methylation spread has been associated with a change in chromatin conformation leading to a repressed state of gene expression .
It is therefore apparent that improved transgene expression can be made to DNA vectors by removal of extraneous bacterial sequences. Yew et al. succeeded in substantially depleting the CpG content of a plasmid vector by successive rounds of site-directed mutagenesis . The reduction of CpG conferred reduced toxicity as well as higher and long-lasting transgene expression of the plasmid compared to a standard CpG replete plasmid in immune-competent mice. This study clearly showed that reduction of CpG was beneficial, but also demonstrated that even a single CpG in the bacterial backbone is sufficient to elicit an inflammatory response.
Several strategies have been described for the production of minimally sized DNA vectors or minicircle DNA, which lack the entire bacterial backbone [13–15]. We have developed a procedure for minicircle DNA production, which utilises Cre recombinase-mediated intramolecular recombination between direct repeats of 34 bp (loxP sites), flanking an expression cassette in a minicircle producer plasmid. The Cre recombinase is expressed under tight metabolic control in the bacterial production strain MM219Cre . When the two flanking loxP sites have the same orientation, Cre site-specific recombination produces two circular and supercoiled DNA molecules, which are topologically unlinked, each containing a single 34-bp loxP site. The two resulting supercoiled DNA circles are the ‘miniplasmid’, which comprises the bacterial backbone, and the ‘minicircle’ comprising the mammalian expression cassette. Since the Cre-recombination reaction is a bidirectional process, we constructed the minicircle producer plasmid with mutant loxP66 and loxP71 sites, which shift the kinetics of the Cre–loxP reaction from equilibrium towards increased minicircle production. The recombination reaction generates a minicircle, containing a hybrid loxP site with impaired recombination ability and a miniplasmid with a wild-type loxP site.
Minicircle DNA vectors have also been produced by other recombinases, such as ϕC31 , λ integrase  and Flp recombinase [16, 17], and each of these minicircle vectors has shown significantly prolonged, transgene expression compared to the mother plasmid, which contains a bacterial backbone. Minicircle vectors have been used in vivo to express reporter proteins , proteins protecting against radiation effects (human manganese superoxide dismutase)  and human proteins such as α1-antitrypsin .
In addition to removing the bacterial backbone, several mammalian DNA elements have been included in plasmid DNA constructs to maximise gene expression and prevent silencing. It is generally accepted that the tissue-specific transgene expression is regulated by a tight interplay between promoter, transgene and additional transcriptional sequences including ubiquitous chromatin opening elements, enhancers, insulators and scaffold/matrix attachment regions (S/MARs) (for review, see ).
In this paper, we focus on the construction and analysis of the expression profiles of S/MAR-containing minicircles comparing tissue-specific and ubiquitous promoters. S/MARs are 70% A/T-rich endogenous sequences, which are generally found close to promoters, enhancers and origins of replication, suggesting that they may link such regulator regions to the matrix-bound DNA and RNA enzymatic machineries . One of the fundamental properties ascribed to them is their DNA strand separation potential under superhelical stress . This is believed to provide functions such as initiation of replication and transcription augmentation, when included in a closed circular, topologically constrained domain of DNA molecule such as an episome . Stress-induced duplex destabilisation (SIDD) in such a domain depends not only on the local sequence properties of a site (such as thermodynamic stability) but also on how that site competes with all other sites in the domain .
A non-viral plasmid vector pEPI-1, containing an S/MAR from the 5′-region of the human interferon β-gene, was shown to replicate episomally in vitro  presumably by recruiting the respective endogenous replication factors of the host cell such as SAF-A protein . However, establishment as an independently replicating episomal entity is a rare and random event. To achieve this in vitro the pEPI-1 vector requires mandatory a 2-week initial period of biological selection pressure, after which it can continue to replicate episomally and express the transgene in the absence of selection .
In contrast, an S/MAR minicircle version of the pEPI-1 vector based on the SV40 promoter was created using the Flp recombinase [16, 17], which showed increased stability and longer EGFP expression in vitro compared to the pEPI-1 vector containing the bacterial backbone. Even more importantly, this S/MAR minicircle shows episomal maintenance without any initial selection pressure. We have recently shown that long-term transgene expression in the mouse liver can be achieved by an S/MAR plasmid when combined with a tissue-specific promoter (such as AAT promoter) .
Here we report the development of novel S/MAR minicircle vectors generated by Cre recombination, using an exonuclease in vitro for minicircle DNA purification. They contain either a tissue-specific (human AAT) or a mammalian ubiquitous promoter (human UbC) with and without an S/MAR. We provide evidence for their episomal status and long-term transgene expression in vitro without any initial selection pressure. By hydrodynamic delivery of these vectors to the murine liver, we demonstrate for the first time that including an S/MAR element in a minicircle vector significantly enhances and stabilises transgenic expression in vivo compared to both a non-S/MAR minicircle vector and an S/MAR vector containing a bacterial plasmid backbone.
Material and methods
The plasmids pAAT-S/MAR (also known as pLucA1) and pUbC-S/MAR were used for construction of the producer plasmid for our new minicircles. The pAAT-S/MAR (7,639 bps) containing the AAT promoter and S/MAR element  was derived from the pEPI-1 vector (kindly provided by Professor Hans J. Lipps, University of Witten, Germany). The pUbC-S/MAR (UbC promoter; S/MAR element; 8,198 bps) plasmid was kindly provided by Dr. Carsten Rudolph (University of Munich, Germany). Both plasmids and their control constructs (without an S/MAR) were designed to express the luciferase transgene and differ only with respect to promoter and the presence or absence of the S/MAR element. All constructs contain an identical bacterial backbone that is derived from the commercial plasmid pGFP-C1 (Clontech, Mountain View, CA, USA). The plasmids were amplified in Escherichia coli DH10B cells (Invitrogen) and isolated using an Endotoxin free Maxi prep kit (QIAGEN, Crawley, UK).
All restriction enzymes for vector construction and testing were purchased from NEB Biolabs. Our original pMLOX6 minicircle producer plasmid (3,489 bps)  contains a chloramphenicol resistance gene and an EcoRV site flanked by two mutant loxP sites, which allow efficient removal of bacterial plasmid sequences by Cre recombination. The minicircle producer pMLOX6-AAT plasmid was created by digesting the pAAT-S/MAR plasmid with NsiI and insertion of the excised 5,237-bp expression cassette (AAT promoter-luciferase-S/MAR-SV40 pA) into the EcoRV site of the pMLOX6 vector. Similarly, the pMLOX6-UbC minicircle producer plasmid was constructed by insertion of the 5,166-bp human UbC promoter-luciferase-S/MAR-SV40 pA expression cassette, as an MluI-PciI fragment from the pUbC-S/MAR plasmid, into the EcoRV site of the pMLOX6 vector. The pMLOX6-AAT-control and pMLOX6-UbC-control minicircle producer plasmids were derived by removal of S/MAR as a 1,974-bp fragment from these vectors using HpaI. Schematic illustrations of each plasmid can be seen in a Supplementary Figure 1.
Minicircle production and purification
The minicircle was prepared as described by Bigger et al.  with slight modifications. Briefly, pMLOX6-AAT (8,726 bps), pMLOX6-UbC (8,655 bps) and their control non-S/MAR minicircle producer plasmids were introduced by electrotransformation into the E. coli strain MM219Cre, expressing the Cre recombinase under tight control of the araC regulon. AraC protein up- or downregulates the expression of Cre depending on the carbon source in the growth medium. In the presence of arabinose transcription from the bad promoter is turned on, whereas in its absence, transcription proceeds at a very low level. Moreover, addition of glucose to the growth medium lowers the levels of 3′, 5′-cAMP, further downregulating the repressed pBAD promoter.
Cells were grown overnight at 30°C in LB medium supplemented with 0.5% glucose and 30 μg/ml chloramphenicol. After washing, cells were resuspended in M9 medium supplemented with 0.5% l-arabinose and grown at 37°C for a further 20 h, instead of the 4–6-h incubation suggested by Bigger et al.  allowing almost complete recombination, thus increasing minicircle yield. After recombination, the different supercoiled DNA forms were isolated from the lysate, and the DNA was digested overnight with FspI (400 U of enzyme/mg DNA), which cuts in the bacterial backbone of pMLOX6 but not in the minicircle; this linearises the producer plasmid and the miniplasmid, but not the minicircle. The undigested minicircle was then purified by CsCl density gradient ultracentrifugation with propidium iodide at 45,000 rpm for 48 h on a Beckman Coulter Optimax centrifuge. Propidium iodide was removed by running the minicircle DNA through a cation exchange column AG50W-X8 (Bio-Rad), and removal of CsCl was achieved either by dialysis, or by adding three volumes of water followed by ethanol precipitation of minicircle DNA and dialysis overnight with water. In a further purification step, ATP-dependent nuclease (Plasmid-Safe nuclease, Epicentr/Biozym Scientific GmbH) was applied to the dialysed minicircles to remove linear or nicked contaminants, followed by recovery of the supercoiled minicircle using the QIAquick PCR purification kit and elution in water according to manufacturer’s protocol. Purified minicircle was then quantified using both a Nanodrop ND1000 Spectrophotometer (Thermo Fisher Scientific, USA) and 0.8% agarose gel electrophoresis. The sizes of the purified minicircles are 5,359 bp for mini-AAT-S/MAR, 5,288 bp for mini-UbC-S/MAR, 3,385 bp for mini-AAT-control and 3,314 bp for mini-UbC-control.
Calculation of SIDD profiles
The analysis of the stress-induced duplex destabilisation profile was performed using the WebSIDD software (freely accessible at www.genomecenter.ucdavis.edu/benham/), based on the technique previously developed by Benham et al. . This method calculates the statistical mechanical equilibrium distribution of a population of identical, superhelical DNA molecules among all available states of denaturation. Consequently, it evaluates the equilibrium probability (p) of denaturation at single base pair (x) resolution and the energy G(x), called also destabilisation energy, which is the incremental free energy required for the denaturation of the base pair at position x. The calculation of G(x) makes it possible to detect fractionally destabilised sites (locations that have a high probability of denaturation), and duplex opening can be driven by relatively small incremental energy. Fractional destabilisation can be biologically important, as it may render a site susceptible to opening by some other process, such as protein binding.
Cell culture—in vitro studies
Human glioma cell lines U251 (ATCC, Rockville, MD, USA) were seeded in 24-well plates and cultured in Dulbecco’s modified Eagle’s medium containing 10% foetal bovine serum and 100 μg of streptomycin/ml in an 5% CO2 incubator set at 37°C. All reagents were purchased from Gibco.
Cells were transfected at 90% confluence using equimolar (0.148 pmol) amounts of mini-UbC-S/MAR (0.51 μg DNA/well), mini-UbC-control (0.32 μg DNA/well) and pUbC-S/MAR (0.8 μg DNA/well) vectors, complexed with Lipofectamine 2000 (Invitrogen) in OptiMEM medium (Gibco) according to manufacturer’s protocol.
After transfection, U251 cells were transferred in 100 × 20 mm tissue culture plates (BD, Falcon) and imaged at regular intervals for bioluminescence using the IVIS Imaging 50 Series (Xenogen) and 150 μg/ml d-luciferin (Gold Biotechnology, USA). Individual colonies were picked and cultured further.
A sandwich enzyme-linked immunosorbent assay (ELISA) was used to measure anti-luciferase antibody levels in animal sera. MaxiSorp™ immunoplates (96-well; Nalge Nunc, Rochester, NY, USA) were coated with 200 μg/ml recombinant luciferase protein (Promega) in 0.1 M carbonate buffer, pH 9.6, and incubated overnight at 4°C (100 μl/well). The plates were washed three times with phosphate-buffered saline (PBS)/0.05% Tween 20 and blocked in PBS/1% milk for 1.5 h at room temperature. Primary antisera were diluted 1:1,000 in PBS/1% milk and incubated in the plates for 2 h at room temperature in triplicate. The plates were washed as above, and a secondary peroxidase-conjugated antibody (DAKO) was added. The plates were then washed five times and detected with the 3,3′,5,5′-tetramethylbenzidine liquid substrate system for ELISA (Sigma). The reaction was stopped by the addition of H2SO4 (0.5 M final concentration). Standards were generated using mouse monoclonal anti-luciferase antibodies (Sigma) with the highest concentration at 100 ng/ml and six serial 1:10 dilutions. Plates were read at 450 nm using a microplate reader.
Southern blot analysis
Total U251 cellular DNA or total liver DNA was extracted using a DNeasy mammalian genomic DNA kit (QIAGEN). The isolated DNA was quantified using a NanoDrop ND1-1000 spectrophotometer (LabTech). Total cellular (10 μg) or liver DNA (30 μg) was digested with or without a non-cutter (EcoRV) and with a single cutting restriction enzyme (HindIII for all AAT promoter and SpeI for all the UbC promoter-based vectors), separated on 0.8% agarose gels (20 V, 20 mA overnight) and blotted onto nylon membranes (Hybond XL, Amersham Biosciences). A 400-bp DNA fragment derived from an EcoNI restriction digest of a segment of the luciferase region, which is common to all plasmids, was labelled with 32P (Rad-Prime labelling kit, Invitrogen) and used as a probe. The hybridization was performed in Church buffer (0.25 M sodium phosphate buffer, pH 7.2, 1 mM EDTA 1% BSA 7% SDS) at 65°C for 16 h.
Plasmid rescue experiments
DH10B E. coli cells were transformed by electroporation, using 1 μg DNA prepared by total liver DNA isolation. Transformed colonies were selected on agar plates containing 30 mg/ml kanamycin. DNA was isolated from individual resistant clones, subjected to restriction analysis (StuI) and analysed by electrophoresis on 0.8% agarose gels.
Comparison of luciferase expression from all constructs was analysed by one-way ANOVA to assess statistical significance. A post-ANOVA multiple comparison procedure (Tukey’s HSD) was further performed to assess pairwise differences on expression confirmed by ANOVA with a significance level p = 0.05.
Animal studies—hydrodynamic delivery
Four-week-old MF-1 (B&K Universal Ltd., UK), each weighing approximately 20 g, was given 2.0 ml PBS containing the plasmid vector DNA or minicircle by hydrodynamic delivery via the tail vein, using a 27-gauge needle. Animals were given adequate care in compliance with institutional and UK guidelines.
For the UbC promoter-based vectors, approximately 0.45, 0.28, 0.7 and 0.53 μg of DNA, for the mini-UbC-S/MAR, mini-UbC-control, pUbC-S/MAR and pUbC-control plasmids, respectively, corresponding to 0.13 pmol of DNA was delivered separately to each mouse, whereas an equal molar amount (1.39 pmol) of mini-AAT-S/MAR, mini-AAT-control, pAAT-SMAR and pAAT-control plasmids, corresponding to 4.9, 3.1, 7 and 5.18 μg of DNA, respectively, was delivered separately to each mouse for the AAT promoter-based vectors.
In vivo bioimaging
At 24 h and in regular intervals after hydrodynamic injections, mice were dosed intraperitoneally (i.p.) with 300 μl of d-luciferin (Gold Biotechnology, USA; 15 mg/ml in PBS), anaesthetised by isoflurane and then imaged for bioluminescence by the IVIS Imaging 50 Series (Xenogen). The Xenogen system reported bioluminescence as photons per second per square centimetre per seradian (sr) in a 2.86-cm-diameter region of interest covering the liver. Data were analysed using Livingimage 2.50 software (Xenogen). Background levels of bioluminescence were 1 × 106 photons/s/cm2/sr. For the partial hepatectomy bioimaging, we ensured that mice were imaged for luciferase from all sides, since liver mass topology can change slightly during post-hepatectomy hepatocyte proliferation.
The amount of vector DNA in transfected hepatocyte samples was calculated by real-time PCR using an ABI PRISM 7000 sequence detector. PrimerExpress™ software was used to design oligonucleotide primers (Invitrogen), the TaqMan probe (MWG) for luciferase to determine amounts of S/MAR plasmid, and primers and probes specific for the mouse Titin gene  were applied to enable normalisation between the samples through calculating the number of cells used as the input. The primers for the luciferase gene were forward: 5′-ggcgcgttatttatcggagtt-3′ and reverse: 5′-ccatactgttgagcaattcacgtt-3′. The probe sequence was 5′-FAM-tgcgcccgcgaacgacatttataat-TAMRA-3′. Amplification reactions (25 μl) contained 5 μl template DNA, 12.5 μl Platinum® Quantitative PCR Supermix-UDG with Rox (Invitrogen), 0.1 mM primers and 0.2 mM probe. Following initial steps at 50°C (2 min) then 95°C (10 min), PCR was carried out for 40 cycles of 95°C (15 s) then 60°C (1 min). Serial dilutions of plasmids containing appropriate sequences to produce a standard amplification curve for quantification and all samples were tested in triplicate.
Minicircle production through Cre site-specific intramolecular in vivo recombination
Long-term expression of replicating S/MAR minicircle episomes in U251 cells
An initial 2-week period of selection pressure was shown by Nehlsen et al. to be an absolute requirement to reveal stable episomal replication of S/MAR plasmids . In that study, the S/MAR plasmid became established in 1% of transfected cells following G418 antibiotic selection and was then able to replicate episomally, without further selection.
Enhanced transgene expression from the S/MAR minicircles in the liver in vivo
Luciferase expression from the mini-AAT-S/MAR and pAAT-S/MAR (Fig. 3a) was similar at day 1, consistent with delivery of equimolar amounts of DNA and therefore the equivalent number of expression cassettes. However, expression mediated by mini-AAT-S/MAR increased approximately 10-fold during the first week and remained significantly higher than pAAT-S/MAR throughout the duration of the experiment (p < 0.001). This clearly demonstrates the considerable effect of removing the bacterial backbone from these plasmids on long-term luciferase expression. At the final time point, expression from mini-AAT-S/MAR was about five-fold its original level at 24 h post-administration. Such an increase of luciferase levels was not observed from any plasmid vector, and levels mediated by mini-AAT-S/MAR remained substantially higher than those expressed by all other vectors, both at day 32 (F = 82.1, p < 0.001) and at the termination of the experiment at 92 days post-administration (F = 39.6, p < 0.001). When directly compared to pAAT-S/MAR, which shows a drop of luciferase expression, mini-AAT-S/MAR mediated a higher level of expression of approximately two orders of magnitude at the final time point (p < 0.05). The benefit of deletion in the bacterial backbone was also evident in the non-S/MAR vectors. In the case of mini-AAT-control, at 2 weeks following administration, luciferase expression was measured at 5% of its level at 24 h post-administration whereas the pAAT-control had dropped to 0.4% of its original value at the same time point (p = 0.05). Towards the end of the experiment, mini-AAT-control remained around 64% higher than the pAAT-control. Significantly, the mini-AAT-control is unable to mediate expression levels above that of either of the S/MAR-containing vectors. Plasmid pAAT-control is quickly silenced and confers only 0.5% of its initial expression at 24 h post-hydrodynamic delivery (p = 0.01).
Similar results were obtained for the UbC-driven vectors. Again longitudinal luciferase expression of the mini-UbC-S/MAR minicircle was in every case higher and more persistent than that of the other constructs over the experimental period. We observed an approximate three-fold increase in luciferase expression from mini-UbC-S/MAR during the first 2 weeks following delivery that remained consistently higher than its initial level throughout the experiment (146% of its original expression value 24 h post-administration at day 92). Mini-UbC-control also sustained a stronger expression level compared to pUbC-control (p = 0.001 at 2 weeks and p < 0.005 at 92 days) further confirming the advantage of removing the bacterial backbone.
In order to evaluate if any of the differences in luciferase transgene expression is related to an immune response, we performed an anti-luciferase ELISA in serum of all treated mice at 92 days post-hydrodynamic delivery. Results showed that in all cases, a near-background level of anti-luciferase antibodies was measured (700 ng/ml) for all treated mice (data not shown). This further indicated that the difference in gene expression is not due to differences in the immune responses against the luciferase transgene.
To investigate the integration status of our AAT and UbC promoter plasmid constructs in the transfected hepatocytes, we isolated DNA from the liver tissues at 3 months post-administration. Total isolated liver DNA (30 μg) was digested with restriction enzymes, which cut the vectors once, either HindIII (for AAT-based vectors) or SpeI (for UbC-based vectors), and was then subjected to Southern blot analysis. A representative blot can be seen in Fig. 4 and all plasmids showed an individual band of the expected size of linear vector in all lanes.
This restriction test cannot, however, exclude the possibility of integrated concatemeric vectors so further analysis was performed to demonstrate that the minicircle vectors are maintained episomally. In this experiment, total liver DNA was isolated as before and digested with EcoRV an enzyme which does not cut either of the minicircle vectors alone or with a the single cutting enzymes used previously. Analysis of EcoRV digested DNA indicated that the minicircle DNA remained intact and migrated as single-copy or multiple-copy aggregates (Fig. 4a lanes 3 and 4 for AAT-based vectors, b lanes 3 and 4 for UbC-based vectors), which where digested into a single full-length linear molecule following the double digestion with HindIII or SpeI (Fig. 4a lanes 5 and 6 for AAT-based vectors, b lanes 5 and 6 for UbC-based vectors). This verifies that our minicircle vectors and their control plasmids did not integrate during the experiment. Hence, all vectors remain episomal at 3 months following administration even though the levels of luciferase expression had declined to near-background levels in some cases.
In a final experiment, we performed plasmid rescue experiments in mice treated with plasmids (pAAT-S/MAR, pUbC-S/MAR and their non-S/MAR controls), using total DNA from the livers of hydrodynamically treated mice. In every case, DNA from the isolated liver sample produced transformed bacterial colonies on plates containing kanamycin, the resistance marker present on all four injected plasmids that contain a bacterial backbone. The restriction patterns of these plasmids were consistent with the unmodified non-integrated plasmid constructs (data not shown). Based on these data, we surmise that silencing and not shedding is the reason for the substantial decline in longitudinal DNA-mediated expression over the 3 months after administration. As the minicircle vectors lack an antibiotic resistance gene, no such rescue analysis could be performed.
Investigation of minicircle replication after partial hepatectomy
Given the difference in expression profiles of the mini-AAT-S/MAR and mini-UbC-S/MAR compared to their non-S/MAR controls and to their plasmid vectors containing the bacterial backbone, we investigated if the minicircle S/MAR vectors had become established as actively replicating episomes in the hepatocytes or if they were only passively maintained and would therefore be lost during hepatocyte turnover.
For this experiment, groups of mice (n = 4) injected with either mini-UbC-S/MAR, mini-UbC-control or pUbC-S/MAR pDNA were forced into rapid liver regeneration by 70% hepatectomy 92 days after administration of the pDNA constructs. Luciferase expression levels were then examined daily for up to 32 days after resection by in situ bioluminescent analysis. Liver regeneration after hepatectomy occurs normally by a process in which remaining hepatocytes undergo one or two cell division cycles until the initial liver mass is reconstituted (after approximately 14 days). During liver regeneration, any actively established vector should replicate and spread throughout the reconstituted tissue leading to a persistence of transgene expression while passively maintained pDNA would become progressively lost. Similar to our observation in a previous experiment with S/MAR plasmid constructs , resection and full regeneration of murine liver led to a reduction of luciferase expression to almost background levels and indicating the passive episomal state of the minicircle constructs (data not shown). Essentially, identical results were also obtained when mini-AAT-S/MAR, mini-AAT-control and pAAT-S/MAR pDNA were used in this experimental setting (data not shown).
Increased domain decondensation of S/MAR minicircles
The luciferase gene has no propensity to separate strands in contrast to the transcription termination sites, which are highly destabilised. It is interesting to note that in both the pAAT-S/MAR (Fig. 5a) and pUbC-S/MAR (Fig. 5b) plasmids, the presence of an S/MAR element creates an area of high destabilisation, which is not present in the control plasmids. Intriguingly, areas of destabilisation present in mini-AAT-control are not present when the vector is linked to the bacterial backbone. These areas correlate with the transcription termination site of the luciferase gene as well as sites of the AAT promoter, which indicates potential binding sites for transcription factors. Direct comparison of the SIDD profile of pAAT-S/MAR and mini-AAT-S/MAR (Fig. 5a) shows that removal of the bacterial backbone created an area of high destabilisation in the AAT promoter area of the mini-AAT-S/MAR vector, not seen in the pAAT-S/MAR plasmid. This becomes more profound in the comparison between the pAAT-control and the mini-AAT-control plasmids, where the removal of the bacterial backbone destabilises the mini-AAT-control plasmid, which could explain the differences in the in vivo higher and prolonged expression of luciferase of the mini-AAT-control plasmid. Similar results are obtained for the UbC promoter.
The therapeutic potential of gene transfer is directly related to the level and persistence of expression of a therapeutic gene and the toxic consequences to the cell, tissue or organism. Non-viral episomal vectors are the most suitable alternative to viral vectors as they do not integrate into the host genome and can be constructed without sequences that might cause cell transformation or toxicity. However, transient gene expression from non-viral vectors by vector-mediated transgene shutdown often due to bacterial DNA sequences is usually the main limitation of these vectors.
Previous studies [7, 14, 15] have shown that transgene expression levels in minicircle vectors are higher and more persistent compared to conventional plasmids. The silencing effects observed in most plasmids have been ascribed to the presence of CpG dinucleotides of pDNA complexed with liposomes and/or the physical linkage of the expression cassette to the bacterial backbone of the pDNA, which typically contains the origin of replication and antibiotic resistance genes . In a very elaborate study using CHiP analysis, Riu et al. showed that the silencing of the transgene from the parental plasmid vector in the liver is accompanied by an increase in heterochromatin-associated histone modifications (H3K9 and H4K20 methylation) and an accumulation of proteins found in heterochromatin domains (HDAC2, HP1 and SUV39H1), as well as by a decrease in modifications typically associated with euchromatin . The pattern of histone modification on minicircle DNA is consistent with characteristic features of active chromatin, such as high concentrations of euchromatic histone modifications (H3/H4 acetylation and H3K methylation). We agree with the hypothesis that transgenic silencing is dictated by chromatin structure rather than directly by specific DNA sequences as stated by Chen and Woo . But it is clear that the composition of a vector clearly has an influence on its chromatinisation. Other studies have described that silencing is not due to direct inhibition of the expression cassette but occurs first at the bacterial DNA sequence before spreading to the promoter and causing its inactivation [8, 29]. Indeed, Chen and Woo have shown that flanking an expression cassette with a chicken HS4 insulator is able to partially overcome silencing from a bacterial DNA sequence but expression from a bacterial backbone free linear expression cassette was still superior . Based on this information, we anticipated that the combination of an S/MAR element in our minicircles and a tissue-specific promoter could both insulate the promoter from the spread of heterochromatin and prevent gene silencing. The use of an S/MAR instead of a cHS4 insulator is in theory better as it was shown to provide better insulation , which may partially explain why our S/MAR vectors show higher gene expression than the no S/MAR vectors. However, both studies highlight the fact that gene expression in the mouse liver can be significantly affected by the combination of regulatory elements such as a tissue-specific promoter with insulating elements (such as the S/MAR or cHS4) that can generate the necessary chromatin structure for long-term luciferase expression.
Consistent with this theory, we generated monomeric supercoiled S/MAR minicircles based on the previously described Cre/loxP recombination system  and using an exonuclease for DNA purification. We found that S/MAR minicircles are indeed established in vitro in the absence of selection pressure and are maintained as episomes, while in vivo adding the S/MAR element to liver-targeting minicircles provides an enhancement to expression and duration, not previously seen with any other vector.
Similar results in vitro have been shown by Nehlsen et al. using S/MAR minicircles generated by the Flp recombinase . This demonstrated that CHO-K1 cells transfected with an SV40 promoter-based S/MAR minicircle were able to express EGFP for at least 55 days, in the absence of antibiotic selection . In this case, EGFP fluorescent cells were picked by FACS sorting 5 days after transfection and were then cultured in the absence of any form of selection, resulting in a population of cells of which 65% were stably expressing EGFP from 12 days after transfection onwards for at least 55 days. This was a significant improvement compared to the original S/MAR plasmid (pEPI-1), where the group reported only 3% of EGFP-positive cells, after 12 days, and even lower at 55 days, in the absence of initial selection pressure. Importantly, this SV40-S/MAR minicircle was found to be episomally maintained during the experiment as shown by Southern blot analysis.
Our results verify and extend these findings in the glioma U251 cell line, by use of a ubiquitous mammalian promoter instead of the SV40 viral promoter used in their studies. In both cases, however, the critical component for long-term transgene expression seems to be the S/MAR element, which is located downstream from the transgene. It is believed that the S/MAR element works by relieving the negative supercoiling that arises during transcription and, by interacting with chromosomal rearrangement factors, protects from gene silencing, thereby allowing long-term episomal vector-mediated transgene expression. The same vectors without an S/MAR element were unable to support long-term gene expression in our in vitro studies. Most importantly, we demonstrate here that these effects can also be achieved in vivo in the adult mouse liver. In clear contrast to the mini-AAT-control vector, our mini-AAT S/MAR vector mediated a three-fold increase in luciferase levels during the first week following hydrodynamic administration that is sustained for the 3-month duration of the experiment. Furthermore, luciferase expression was sustained approximately 10-fold over that of the ubiquitous mini-UbC-S/MAR and roughly 20-fold over that of the previously published pAAT-S/MAR .
When comparing expression between mini-AAT-control and pAAT-S/MAR, no significant differences exist suggesting that the presence of the S/MAR is able to offset transcriptional silencing mediated by bacterial sequences in the plasmid backbone perhaps by a combination of histone modifications associated with transcriptionally competent chromatin and prevention of variegation . This is also observed when comparing mini-UbC-control and pUbC-S/MAR. In contrast, the difference in luciferase expression between mini-UbC-S/MAR and pUbC-S/MAR in MF1 immunocompetent mice is significant, confirming that S/MAR is unable to prevent silencing on a non-specific promoter even when the frequency of unmethylated CpG motifs was reduced.
It should also be mentioned that expression from the luciferase transgene itself in mice has been shown to be immunogenic before . We performed an anti-luciferase ELISA on blood serum of our MF1 treated mice at 92 days after hydrodynamic injection, but we could not observe a significant antibody response, with levels of anti-luciferase antibodies being similar for all treated mice (at approximately 700 ng/ml). We believe that this could be the result of the extremely low quantities of DNA vector used. In support of that, there is a report where a single hydrodynamic delivery of 10 μg plasmid DNA with either the UbC promoter driving expression of luciferase, produced significant anti-luciferase antibodies above 14,500 ng/ml . Crucially, we show that our vectors remain episomal in vivo, with no evidence of integration, supporting previous studies that showed passive episomal maintenance in the mouse liver, of circular DNA molecules [9, 34, 35].
This paper demonstrates for the first time the significant enhancement of gene expression and maintenance of episomal vector-mediated transgene expression by minicircle vectors based on S/MAR in vivo. A previous study from Jacobs et al.  has reported that transgene expression from a minicircle vector containing an S/MAR element peaked between day 7 and day 10 after injection and remained fairly stable for at least 35 days. However, these results are likely to be attributed to the inclusion of several additional enhancer elements. This is in contrast to our study where we directly compare the effects of the presence of the S/MAR element in combination with different promoters, for an extended period of time. However, despite the potentially life-long transgene expression from our mini-AAT-S/MAR minicircle in normal liver, this vector did not show any detectable episomal replication as demonstrated after induction of cell division by partial hepatectomy. This finding is in accordance with our previous in vivo results with the pAAT-S/MAR plasmid  and verified in another study using another minicircle DNA system  but differs from our in vitro results, where the minicircle vector is episomally established and propagated without selection pressure after isolating a strongly expressing colony of cells. We recently published a report  suggesting that episomal replication conferred by the S/MAR in vivo does occur at a low level but requires some form selection mechanism to reveal it. Similarly in vitro, the expansion of S/MAR minicircle transfected cells from an isolated colony, without competition from untransduced cells, would provide a certain proliferative advantage even without antibiotic selection.
It is generally accepted that the tissue-specific transgene expression is regulated by a tight interplay between promoter, transgene and additional transcriptional sequences, based on previous reports that show that transcription initiation rates from an expression cassette combined with an S/MAR element are increased , as well as the ability of the S/MAR to prevent expression silencing in vivo, at least when combined with a tissue-specific promoter . We may, therefore, speculate that the following molecular events contribute to the observed prolonged luciferase expression from our S/MAR minicircle vectors compared to their bacterial backbone containing plasmids: (1) histone modifications associated with transcriptionally competent chromatin, (2) increased molecular destabilisation to mediate domain opening (based on the SIDD profile studies) and (3) the ability of S/MAR to shield promoter methylation, efficiently maintaining transgene expression . Loosening the chromatin and DNA structure gives access to transcription factors and/or to the origin recognition complex. For example, it has been reported that S/MAR binds SAF-A (a major RNA binding protein that recruits active genes to the nuclear matrix by binding the transcriptional coactivator p300) in vivo, which results in increased transcription initiation rates . Our results demonstrate the complex interplay of gene expression between different vector elements and adoption of an appropriate chromatin structure.
It is interesting to note that luciferase expression from both the mini-UbC-S/MAR and mini-AAT-S/MAR vectors steadily increased during the first week after hydrodynamic delivery. It is possible that the S/MAR minicircles adopt a stable open chromatin status, which sustains long-term luciferase expression by the end of the first week. This possibility is currently under investigation. At least for the UbC promoter (albeit without an S/MAR element), this phenomenon has been reported before  and has been partially attributed to the several sp1 transcription factor binding sites present on the UbC promoter, which are associated with histone acetyltransferases.
In conclusion, an ideal non-viral vector for gene therapy should be non-toxic, have mitotic stability, allow non-integrative establishment and provide persistent therapeutic expression levels. The S/MAR-containing minicircle vector that we have developed fulfils these requirements. Future studies will investigate the exact mechanism of action of the S/MAR element to support long-term gene expression in vivo. It is reasonable to believe that such vectors could be easily modified to contain an S/MAR element with a tissue-specific promoter for use in other tissues apart from the liver.
The work was supported by the Myrovlytis Trust.
Conflict of interest
The authors declare no conflict of interest.