Advances in Non-Viral DNA Vectors for Gene Therapy

Uses of viral vectors have thus far eclipsed uses of non-viral vectors for gene therapy delivery in the clinic. Viral vectors, however, have certain issues involving genome integration, the inability to be delivered repeatedly, and possible host rejection. Fortunately, development of non-viral DNA vectors has progressed steadily, especially in plasmid vector length reduction, now allowing these tools to fill in specifically where viral or other non-viral vectors may not be the best options. In this review, we examine the improvements made to non-viral DNA gene therapy vectors, highlight opportunities for their further development, address therapeutic needs for which their use is the logical choice, and discuss their future expansion into the clinic.


Introduction to Gene Therapy
Gene therapy is the use of nucleic acids to repair, replace, or regulate genes to prevent or treat disease [1]. Hundreds of genes have been investigated as potential gene therapy candidates. Some notable examples include replacement of the mutated cystic fibrosis transmembrane conductance regulator with a functional copy to treat cystic fibrosis [2], the knockdown of C-C chemokine receptor type 5 (CCR5) to block cell entry by human immunodeficiency virus (HIV) [3], the expression of glucagon-like peptide 1 mimetics to regulate blood glucose levels in diabetic patients [4], and expression of viral antigens in a DNA vaccine for influenza that eliminates the need for potentially active viral particles [5].
Gene therapy vectors are broadly categorized as viral or non-viral [1]. Viral vectors are widely used because of their natural ability to invade cells and deliver a manipulated genetic payload for therapeutic use. It is far more difficult for non-viral vectors (RNA or DNA) to transfect many specific cell types, so they are usually complexed with delivery vehicles (e.g., cationic lipids, cationic polymers, etc.) or subjected to forced entry (e.g., electroporation, hydrodynamic injection, etc.). Advancements in transfection methods are occurring rapidly; for an overview of these delivery technologies, we refer the reader to the following reviews [6][7][8][9].
Lentiviral and retroviral vectors, which are designed for insertion into the genome, bring a high risk of gene disruption [10,11]. Adeno-associated virus (AAV) vectors have a lower but still existent risk of insertional mutagenesis [12]. Adenoviral vectors are maintained episomally, which is advantageous [13], but can cause toxicity and immunogenicity [14]. Certain viral vectors, such as those specifically based on adenovirus serotype 5 (AdV5) or adeno-associated virus type 2 (AAV2), cannot be used because the virus is so widespread that many people have a pre-existing immunity [15]. In all of these families of viral vectors, even those with low seroprevalence in the human population, the same serotype of construct cannot be delivered repeatedly to the same patient because they will have developed an immunity to it [12,16,17].
Non-viral vectors are far less immunogenic than viral vectors [18]. RNA delivery for RNA interference (RNAi)-short interfering RNAs (siRNA), microRNA (miRNA), etc.-, however, comes with the difficulty of expense and scale because they turn over so quickly. Longer RNAs, such as synthetic messenger RNAs (mRNAs), which are delivered to the cell and translated in vivo, offer potential for expressing proteins. The nucleotides in these mRNA molecules must be modified to avoid immune detection through pattern recognition receptors such as Toll-like receptor 3 (TLR-3), TLR-7, TLR-8, and retinoic acid-inducible gene I (RIG-I) [19]. RNA vectors are typically less stable and more transient than DNA and thus require additional protecting measures (end-blocking, base modification, vehicle choice, etc.) [20]. These protection measures, however, may introduce their own confounding issues.

Challenges of Using Plasmid Vectors for Gene Therapy
Plasmid vectors for gene therapy are beset with some notable inherent limitations. Most plasmid DNA preparations contain several topological variants of the plasmid, including supercoiled (the preferred topology), but also the unwanted open circular and linear forms of the molecule. As mentioned above, plasmids are generally inefficient at delivering their payloads compared to viruses, thus requiring vehicles, physical forces, or specialized modifications for uptake and nuclear localization [25,26]. Some of these delivery methods lead to breakage of plasmid DNA backbone, which increases the likelihood of genome integration and, if the break occurs in the therapeutic sequence, less efficient expression [27].
Because plasmids are non-replicating episomes, transgene expression is transient and diluted by cell division. Additionally, bacterial sequences in plasmids can contribute to their gene silencing [28,29]. Unmethylated cytosine-phosphate-guanine (CpG) dinucleotides, which are more common in bacterial DNA than in mammalian DNA, have the potential to be recognized by the mammalian immune system via TLR-9, potentially precipitating not only transgene silencing but also immune response [30].
Plasmids contain a bacterial origin of replication (ori) for propagation in a bacterial host strain. An ori potentially allows plasmids meant to deliver therapeutic sequences to also inadvertently transfer into and replicate in other bacteria. Additionally, plasmids encode genes (typically antibiotic resistance-encoding genes) for selection of plasmid-harboring bacteria. The use of antibiotics and their resistance genes in the preparation of plasmid vectors, however, is discouraged by regulatory bodies such as the Food and Drug Administration and the European Medicines Agency because of the risk of transfer and replication of resistance genes to bacteria in the human microbiome and possibly into the environment. Additionally, residual antibiotics that remain from vector production may trigger an immune reaction in patients.
Because of the challenges, extensive modifications have been made to plasmids to satisfy regulatory requirements for clinical use in humans [18]. These modifications involved the deletion of unwanted and unnecessary sequences and have resulted in the advent of minimized DNA vectors.

Improvements to Plasmid Vectors
Several studies have revealed that decreasing plasmid size improves transfection efficiency, suggesting that minimizing vector length should be one of the goals of non-viral vector design [31][32][33]. Attempts to minimize and otherwise optimize plasmid DNA vectors for gene therapy first involved the removal of antibiotic resistance genes [34][35][36], thereby increasing clinical potential. Because of this removal, however, antibiotic-free systems had to utilize a different mechanism for selection. One of the first systems was operator repressor titration (ORT) [34]. ORT plasmids (pORT) contain one or more operator sequences that are used to titrate, through competition, repressor proteins (e.g., Lac repressor) that normally bind an endogenous operator sequence upstream of an essential, chromosomally encoded gene in bacteria.
In the case of pORT, the essential gene is dapD. Repression of dapD is lethal when cells are grown on specialized growth medium. Bacteria that harbor multi-copy plasmids outcompete the repressor from the endogenous operator and survive. This selectable marker system has only minimal bacterial sequences encoded on the plasmid (ori and short operator sequences), requires no plasmid gene expression, promotes plasmid stability, and, if need be for production, can be used in any microorganism able to propagate plasmids. pORT has been used effectively in pre-clinical and clinical studies as a DNA vaccine [37][38][39] and as a component of live bacterial vaccines using Salmonella enterica [40][41][42].
Other DNA vector selection systems that are free of antibiotic resistance genes rely on chromosomal mutations in special producer strains that are complemented only when plasmid is present. One of these systems is the plasmid with conditional origin of replication (pCOR) and another is the plasmid free of antibiotic resistance (pFAR). Both rely on amber mutations in essential chromosomal genes (the genes encoding arginine and thymidine, respectively), creating auxotrophic bacteria that only grow if they harbor plasmid containing the complementary amber suppressor tRNA.
The conditional origin of replication in pCOR, ori-γ from the R6K class A theta plasmid, relies upon a π initiator protein that is produced only within a narrow host range by the pir gene [43]. This safeguard means that pCOR would be unlikely to disseminate into the environment. Additionally, pCOR may be less immunostimulatory than standard ColE1-derived plasmids [35]. pCOR has demonstrated higher levels of reporter gene activity compared to commercially available plasmids [35] and also has been brought to clinical trial for critical limb ischemia [44]. Similarly, the pFAR system has been used to effect high luciferase expression in both transplanted tumor cells and in the skin of mice [36]. In both cases, pFAR vectors displayed higher gene expression and persistence than standard plasmid vectors.
By decreasing vector length, removal of antibiotic resistance sequences had the additional advantage of improving the efficiency of vector production in many cases because of decreased metabolic burden upon the bacterial host. Removal of these genes also meant that costly and time-consuming procedures for evaluating the amount of residual antibiotics left in vector preparations could be eliminated.
The vectors specifically mentioned above serve as highlighted examples; they and other antibiotic-free miniplasmid systems have been more thoroughly reviewed elsewhere [18,45,46]. In Table 2 we list DNA vectors and their general composition (whether they encode a bacterial origin of replication or an antibiotic resistance-encoding gene). Although we had hoped to include other key features (e.g., minimal and maximal vector length, expression efficiency, etc.) for all of the vectors, data were not consistently available. Therefore, we provide several distinguishing advantages and disadvantages of each.  [35,44,[51][52][53][54] pFAR Vector is only 1.1 kb The amber mutation used is very efficiently suppressed Growth medium does not contain animal-derived components Production process yields mostly monomeric closed circular DNA Both the plasmid and host strain are well-defined and described Mutant host strain required for propagation tRNA expression required for selection [36] Post-segregational killing (PSK) systems Sequence addition to miniplasmid is short Toxin is highly efficient for selection PSK genes can cause mild toxicity [55][56][57] RNA IN/RNA OUT RNA sequence is only transcribed in prokaryotes Mutant host strain required for propagation [58,59] RNA I/RNA II No sequence addition to plasmid necessary Can be produced in gram quantities more easily than standard plasmids Mutant host strain required for propagation [60,61] Overexpression systems No mutant host strain needed Not much shorter than plasmids Potential metabolic burden on host strain Possible antibiotic contamination of final product [62,63]

MiLV
Production process avoids bacterial contaminants such as endotoxin PCR amplification of product is potentially error prone [85] Ministring Chromosomal integration causes apoptotic cell death One-step in vivo production system This system can also create ccc vector topology Uses ampicillin resistance as a selection marker Residual ampicillin or endotoxin possible in final product [86][87][88][89] Mini-intronic plasmid

Yes 2 No
Production process and yield is the same as standard plasmids, but subsequent splicing removes problematic sequences Inclusion of introns is reported to enhance transgene expression Uses RNA IN/RNA OUT as a selectable marker system Reported to have higher transgene expression levels than plasmids or minicircles Vector length is not as reduced as some minimal systems, requiring the use of more transfection reagent Small chance of unspliced or mis-spliced mRNA Possible immune recognition of ori or other bacterial sequences before splicing occurs [90] ori : origin of replication; AR: antibiotic resistance; CpG: cytosine-phosphate-guanine dinucleotide; pORT: operator repressor titration plasmids; pCOR: plasmids with conditional origin of replication; pFAR: plasmids free of antibiotic resistance; tRNA: transfer RNA; PSK: post-segregational killing; MIDGE: minimalistic immunologically defined gene expression; MiLV: micro-linear vector; PCR: polymerase chain reaction; 1 True of most if not all non-viral DNA vectors; 2 Mini-intronic plasmids are produced as standard plasmids and initially contain the same elements when transfection takes place. Only after splicing has occurred inside the target cell are undesirable elements such as ori finally removed.

Minicircles
Despite the removal of some problematic sequences, significant bacterial sequences remain in antibiotic resistance-free systems. As mentioned above, bacterial sequences, and particularly ori, have the potential for triggering inflammation or for silencing transgenes [91]. Removal of these additional extraneous sequences was first reported in 1997 by Darquet et al. [65]. These researchers used site-specific recombination to turn parent plasmids (containing the regulatory sequences necessary for intramolecular recombination) into minicircles, containing the therapeutic sequences desired and only very short segments of bacterial DNA. The rest of the bacterial DNA from the parent plasmid, including ori, is recombined into a discarded miniplasmid. Because minicircles no longer contain ori, they cannot replicate in bacteria and are, thus, no longer considered plasmids.
Multiple recombinase systems have been used to generate minicircles, including phage λ integrase, phiC31 recombinase, Flp recombinase, ParA resolvase, and Cre recombinase (reviewed in [9,18]). In the miniplasmid systems mentioned in Section 3, the miniplasmids are, themselves, the gene therapy vector (because only antibiotic sequences were removed from them), whereas here they are discarded because they contain everything except the sequence of interest. In both cases, miniplasmids encode ori, and are therefore still referred to as plasmids.
Plasmid contaminants in minicircle preparations can be as high as 10% of the total yield-well above the 1.5% allowed by some health regulatory agencies [92]. Because of this problem, several methods for improving minicircle purity have been developed: (i) a triple helix DNA technology (TriD) that uses biotinylated oligonucleotides and streptavidin-coated magnetic beads to selectively remove parent plasmid and miniplasmid [92]; (ii) incorporation of unique nicking endonuclease sites outside of the minicircle sequence on the parent plasmid to facilitate separation of minicircles from contaminating products with hydrophobic interaction chromatography [93]; (iii) the use of an anion-exchange monolithic column (CIM diethylamine) for selective separation of minicircle from the unwanted products [94]; and (iv) enhancing ParA resolvase activity to increase yield during fermentation by boosting recombination for minicircle generation [95]. These methods all work to improve yield of minicircles, but add both time and expense to the preparation. Additional optimization may be needed to generate minicircles on an industrial scale.

Minivectors
Minivectors are minimized, non-viral DNA vectors similar to minicircles but with some important differences. Like minicircles, minivectors are synthesized from a parent plasmid via site-specific recombination ( Figure 1) [73]. Encoding only the genetic payload and short integration sequences, minivectors can be engineered as small as~350 bp and generated in high yields (in comparison, the smallest reported minicircle length is 650 bp [32]; the yield of minicircles this small is unclear). As before, unwanted bacterial sequences are on a discarded miniplasmid. The recombination and purification system used to make minivectors is highly optimized, resulting in as much as 100-fold lower plasmid contamination than is recommended by health regulatory agencies [92]. To generate minivectors, any target sequence or expression cassette is cloned between the attB and attP sites located in direct orientation in a minivectorproducing parent plasmid. λ integrase mediates the intramolecular recombination of the attB and attP sites, producing two catenated rings: the minivector containing the target sequence or expression cassette, and a miniplasmid containing all the other undesired sequences. The catenanes are unlinked by topoisomerase IV. λ Int: lambda integrase; Topo IV: topoisomerase IV; bla: β lactamase (encoding ampicillin resistance); ori: bacterial origin of replication; attP: phage attachment site; attB: bacterial attachment site; attL: hybrid attachment site to the "left" of the recombined sequence; attR: hybrid attachment site to the "right" of the recombined sequence.

Increasing the Functionality of Minimized DNA Vectors
Compared to plasmids, minicircles and minivectors transfect cells better in vitro, ex vivo, and in vivo [31][32][33], express transgenes more effectively [67,96], and have an improved safety profile for clinical use. In addition to having no problematic antibiotic resistance genes and negligible bacterial sequences, the benefits of reducing vector length for gene delivery and transgene expression have been well established [31][32][33]97]. Naked DNA vectors less than ~1,200 bp completely survive aerosolization, a direct consequence of being beneath a sharp threshold hydrodynamic radius [71]. Increased negative supercoiling makes minivectors even more compact, which both promotes nuclear localization and provides additional protection from the shear forces of aerosolization (in addition to the protection afforded by their smaller sizes) [71,98]. There is a dramatic length dependence for naked DNA vectors surviving human serum; the smaller the vector, the longer the survival [72,99]. For these reasons, then, minicircles and minivectors are a good choice for gene therapy trials. Additional modifications that could further improve minimized vectors follow.

Nuclear Localization Signals
For a therapeutic vector to be efficiently delivered, it must first be able to reach the target cells. After that, a number of physical barriers must be traversed-the cell membrane, the cytoplasm, and, finally, the nuclear membrane. This last step is considered the rate-limiting step in transfection because most exogenous naked DNA (purified plasmids with no delivery vehicles) or complexed DNA vectors are too large to passively diffuse through the nuclear membrane [18,100].
Associating nuclear localization signals (NLSs) with non-viral DNA constructs would be useful for therapeutic applications, particularly in non-dividing cells. Found on proteins destined for the nucleus, NLSs are clusters of amino acid "marks" of positively charged residues, like arginine and lysine, that are recognized by karyopherins such as importin α. Karyopherins facilitate transport across the nuclear envelope through nuclear pore complexes. Plasmids mixed with cytoplasmic shuttle proteins containing NLSs are shuttled across the nuclear membrane, although at a rate at least two orders of magnitude slower than that of NLS-containing proteins alone [100]. This slow rate is probably a consequence of the time it takes for exogenous DNA and NLS-containing proteins to associate with each other inside the cell. If NLSs are added directly to DNA vectors, time into the nucleus could be reduced and gene expression increased [100]. To generate minivectors, any target sequence or expression cassette is cloned between the attB and attP sites located in direct orientation in a minivector-producing parent plasmid. λ integrase mediates the intramolecular recombination of the attB and attP sites, producing two catenated rings: the minivector containing the target sequence or expression cassette, and a miniplasmid containing all the other undesired sequences. The catenanes are unlinked by topoisomerase IV. λ Int: lambda integrase; Topo IV: topoisomerase IV; bla: β lactamase (encoding ampicillin resistance); ori: bacterial origin of replication; attP: phage attachment site; attB: bacterial attachment site; attL: hybrid attachment site to the "left" of the recombined sequence; attR: hybrid attachment site to the "right" of the recombined sequence.

Increasing the Functionality of Minimized DNA Vectors
Compared to plasmids, minicircles and minivectors transfect cells better in vitro, ex vivo, and in vivo [31][32][33], express transgenes more effectively [67,96], and have an improved safety profile for clinical use. In addition to having no problematic antibiotic resistance genes and negligible bacterial sequences, the benefits of reducing vector length for gene delivery and transgene expression have been well established [31][32][33]97]. Naked DNA vectors less than~1,200 bp completely survive aerosolization, a direct consequence of being beneath a sharp threshold hydrodynamic radius [71]. Increased negative supercoiling makes minivectors even more compact, which both promotes nuclear localization and provides additional protection from the shear forces of aerosolization (in addition to the protection afforded by their smaller sizes) [71,98]. There is a dramatic length dependence for naked DNA vectors surviving human serum; the smaller the vector, the longer the survival [72,99]. For these reasons, then, minicircles and minivectors are a good choice for gene therapy trials. Additional modifications that could further improve minimized vectors follow.

Nuclear Localization Signals
For a therapeutic vector to be efficiently delivered, it must first be able to reach the target cells. After that, a number of physical barriers must be traversed-the cell membrane, the cytoplasm, and, finally, the nuclear membrane. This last step is considered the rate-limiting step in transfection because most exogenous naked DNA (purified plasmids with no delivery vehicles) or complexed DNA vectors are too large to passively diffuse through the nuclear membrane [18,100].
Associating nuclear localization signals (NLSs) with non-viral DNA constructs would be useful for therapeutic applications, particularly in non-dividing cells. Found on proteins destined for the nucleus, NLSs are clusters of amino acid "marks" of positively charged residues, like arginine and lysine, that are recognized by karyopherins such as importin α. Karyopherins facilitate transport across the nuclear envelope through nuclear pore complexes. Plasmids mixed with cytoplasmic shuttle proteins containing NLSs are shuttled across the nuclear membrane, although at a rate at least two orders of magnitude slower than that of NLS-containing proteins alone [100]. This slow rate is probably a consequence of the time it takes for exogenous DNA and NLS-containing proteins to associate with each other inside the cell. If NLSs are added directly to DNA vectors, time into the nucleus could be reduced and gene expression increased [100].
There are several ways in which an NLS can be joined to a DNA construct [101,102]; some methods include electrostatic attraction, covalent linkage via direct chemical conjugation, or complexation of DNA and NLS-containing proteins using various linker molecules. In the latter strategy, Vaysse et al. took advantage of the high affinity of the tetracycline repressor (TetR) for the tetracycline operator sequence (tetO) [103]. They engineered fusions of the tetracycline repressor to both the well-studied SV40 large T antigen-derived NLS (TetR-NLS) and the HIV Tat (TetR-Tat) peptide. Upon sequential intravenous injection [104] of lacZ-expressing minicircles that also contained multiple tetO sequences, this group found a six-fold increase of β-galactosidase expression in murine lungs when minicircles were combined with the TetR-Tat fusion compared to parent plasmid injection alone [103]. in vitro, the transfection efficiency was more than 30-fold higher in A549 cells using minicircle and TetR-NLS compared to plasmid transfection alone [103]. The advances afforded by these proof-of-principle experiments will hopefully translate into the clinic in the future.

Cytosine-Phosphate-Guanine Dinucleotides
A CpG motif contains a centralized cytosine-phosphate-guanine dinucleotide flanked by regions of different length and sequence depending on the type of CpG [105]. Unmethylated CpG motifs from bacterial DNA are pathogen-associated molecular patterns (PAMPs) that stimulate the innate immune response by triggering cells that contain TLR-9, such as human B-cells and plasmacytoid dendritic cells [106]. Members of the TLR family of proteins serve as pattern recognition receptors capable of detecting certain PAMPs. TLR-9, in particular, can detect unmethylated CpG motifs, which are at least four times more common in prokaryotic DNA than in eukaryotic DNA [107]. Thus, many plasmid vectors already contain an inbuilt source of immunogenicity that can be valuable for DNA vaccination.
CpG motifs have been encoded within plasmid DNA vaccines [108,109], or co-delivered with DNA vaccines on either other DNA vectors [110] or oligonucleotides (reviewed in [111,112]). Miniplasmid vectors have demonstrated utility as improved DNA vaccines [37][38][39]. Although minicircles have only been tested pre-clinically in a limited number of tests [113], interest in this application is mounting [27,70]. Minivectors have not yet been examined in this capacity to date. However, as with plasmid vectors above, CpG sequences can be added to minicircles or minivectors or delivered via additional oligonucleotides to promote an immunostimulatory effect when they are delivered as vaccines or co-delivered as adjuvants with other vaccines.

Scaffold/Matrix Attachment Regions
Because minicircles and minivectors exist episomally and do not replicate on their own, their effectiveness dilutes with cell division or cell death. Although this transience is ideal for many therapeutic applications, others require more persistence. One way to circumvent the transient nature of minicircles and minivectors, particularly in quickly dividing cells (e.g., the hematopoietic system), would be to engineer them to contain scaffold/matrix attachment regions (S/MARs) [114,115]. S/MARs are endogenous AT-rich sequences that play an important role in the spatial organization of chromosomes through DNA loop base attachment to the nuclear matrix. S/MARs are often found close to regulatory elements such as promoters, enhancers, and origins of DNA replication [116]. S/MARs can be incorporated into DNA vectors to facilitate a once-per-cell-cycle replication to maintain the vector as an episome in daughter cells [116,117]. To confer function as an episome, an S/MAR sequence must be encoded downstream of an actively transcribed gene [117].
S/MAR sequences are ubiquitous and widely spread in eukaryotic genomes, with 453 in just one 30 Mb region of the human genome [118]. Because they are typically long sequences, averaging~5 kb in length, their addition into minimized non-viral vectors would cause the loss of the advantages of small size. Furthermore, only a few S/MAR sequences have been validated to confer episomal vector maintenance. One such validated sequence, a~2 kb S/MAR from the human β-interferon gene cluster, was successfully used to make a non-integrating lentiviral vector persist in dividing cells [119][120][121][122].
A direct comparison of a S/MAR-containing minicircle to a S/MAR-containing plasmid revealed that 65% of cells stably produced a green fluorescent protein (GFP) reporter 55 days post-transfection with the minicircle compared to only 3% of GFP-positive cells after transfection with the plasmid [123]. Cells transfected with minicircles without S/MAR lost luciferase expression in glioma cells after one week, but S/MAR-containing minicircles maintained expression for two months [117]. Similar results were seen in hydrodynamically injected mouse liver. In a direct comparison of minicircles and plasmids with and without S/MARs, only minicircles containing S/MAR expressed luciferase for 92 days [117]. Despite the length an S/MAR would add to an otherwise small vector, these successes illustrate the potential of S/MAR sequences to confer persistence for long-term gene therapy effects with vectors that otherwise do not replicate.

Viral Replication Genes
The inclusion of viral sequences in episomal vector systems has granted them extended replication in mammalian cells and has proven to be a useful tool for gene expression studies and gene therapy [124]. Epstein-Barr virus (EBV) is a herpes virus that replicates its genome episomally in host cells upon latent infection [125] by expressing Epstein-Barr nuclear antigen 1 (EBNA-1), which in turn recognizes the oriP site and initiates replication [125]. This system was exploited by encoding oriP and EBNA-1 onto a DNA vector for the long-term replication of plasmids in mesenchymal stem cells [126] and human fibroblasts [127]. Other viruses, such as SV40 and papillomaviruses, have similar systems with trans-acting elements that allow them to replicate in mammalian cells [128]. For a more detailed overview of the application of these viral sequences in episomal expression vectors, we refer the reader to articles by Van Craenenbroeck et al. [124] and Jackson et al. [128]. The generation of viral hybrid vector systems was an attempt to circumvent the safety concerns of viral vectors. Although there is potential for minimized vector systems to encode these viral sequences, thus turning them into minimized episomes, inclusion of these sequences would, again, result in increased vector length.

DNA Vaccines
For DNA vaccines, antigens of interest are encoded on DNA vectors in an expression cassette. These vectors allow the expression of foreign proteins and cause subsequent formation of antibodies in the vaccinated host. In spite of the potential advantages afforded by plasmids for this endeavor, their development as DNA vaccine vectors was quickly hindered by their failure to elicit desired levels of immunogenic responses. Their failure was a consequence of poor transfection, transient antigen expression, and transgene silencing [129]. One of the first incentives to create minimal DNA vectors was for use as improved DNA vaccines. Because the presence of bacterial segments of 1,000 bp or more mediates transgene silencing in some tissues (reviewed by Williams [130]), more prolonged and sustained antigen expression can be achieved with minimized vectors than with plasmid vectors [70,130].
Linear contaminants or any closed circular molecules that break upon vaccine delivery increase the probability that DNA fragments may insert into chromosomes disadvantageously. Minimized DNA vectors, which are resistant to shear forces and thus not likely to break, represent the natural next step in vaccine development. In addition, because there are many more molecules of a smaller vector per mass than a larger vector, this means there are a higher number of expression cassettes per volume. Thus, an effective vaccine dose could be achieved in a smaller volume [27].
Three clinical trials have been performed to treat chronic hepatitis B virus (HBV) using a plasmid DNA vaccine expressing the S envelope protein of HBV. Two of these trials resulted in seroconversion in 50% of patients, and the third trial showed that combination therapy of the DNA vaccine with an antiviral drug was less efficacious in people of Asian descent than it was in people of Caucasian descent (reviewed in [131]). This result demonstrated proof-of-principle that plasmids could be used as DNA vaccines, but also revealed plenty of room for vector improvement. Despite numerous pre-clinical studies, only one plasmid vector vaccine has been approved for use in humans at this time [132]. Minimized vectors might ultimately have better utility. For example, Dietz et al. found that minicircles elicit a stronger antigen-specific T-cell response and gave greater protective immunity than plasmids did in a mouse model of listeriosis [97].
DNA vaccines are currently being pursued as a replacement for live attenuated virus vaccines for the seasonal influenza virus. DNA vectors can be administered intradermally, without a needle. One such polyvalent DNA influenza vaccine was recently successfully tested in rabbits and pigs, as is standard progression for influenza vaccine testing [133,134]. In these animal trials, the minimized antibiotic-free plasmid constructs (1,700 bp plus either a 1,778 bp hemagglutinin gene or a 1,413 bp neuraminidase gene [135]) were more efficacious than the traditional plasmid constructs (3,665 bp plus either the hemagglutinin or the neuraminidase gene) [134]. These DNA vaccines could be made even smaller with the minivector system (106 bp plus the antigenic sequences), which should further improve antigen gene delivery and expression.

Cancer and Immunotherapy
Plasmid vectors and, in at least one case, minimized DNA vectors, have been used successfully to deliver gene therapies against cancer (Table 1) [47,[136][137][138]. In pre-clinical work to induce antitumor activity, Wu et al. used minicircles to induce the expression of interferon-γ in nasopharyngeal carcinoma cells. Expression of interferon-γ had a profound anti-proliferative effect in vitro and a survival-mediating antitumor effect in xenografted mice [139]. Minimal-size (MIDGE) vector systems have been used successfully both pre-clinically and clinically as DNA vaccines against advanced stage cancers in conjunction with double stem loop immunomodulator molecules (dSLIM) that serve as effective immune adjuvants [76,140,141]. MIDGE has also been tested for the sensitization of melanoma cells to chemotherapy [142]. Furthermore, in 2011, Zhao et al. showed the potential of minivectors encoding short hairpin RNA (shRNA) against anaplastic lymphoma kinase (ALK) [72]. They demonstrated increased transfection efficiency and gene silencing capability of minivectors compared to plasmid, and equivalence of gene silencing compared to siRNA. They found that minivectors transfect Jurkat cells, which are normally refractory to transfection, and were able to slow the growth of anaplastic large cell lymphoma cells in vitro using knockdown of ALK [72]. Minicircles were also recently used to engineer T-cells with Sleeping Beauty transposon [143] and were further used to deliver bi-specific antibodies, allowing T-cells to kill B-cell lymphomas [144]. Minicircles have also been incorporated into cancer detection systems, where they are used as tumor activators to facilitate detection of endogenous blood biomarkers [145]. Theoretically, minimized DNA vectors also could be used to restore expression of tumor suppression genes, silence transcripts of oncogenic proteins, or sensitize cells of the immune system against malignant cells.

Stem Cell Reprogramming
Minimized DNA vectors can be used to reprogram somatic cells for generation of induced pluripotent stem cells (iPSC). In 2010, Jia et al. created a 2A-linked polycistronic minicircle containing the genes for the reprogramming factors Lin28, Oct4, Sox2, and Nanog, with GFP as a reporter [146]. This group was able to induce pluripotency in human adipose stem cells. Although the rate of reprogramming was low compared to that of viruses, iPSC reprogrammed with minicircle DNA formed embryoid bodies in culture and teratomas in immune-deficient mice. They found minicircle efficiency was higher than that of plasmid DNA [146].
More recently, Daneshvar et al. created iPSCs from umbilical cord mesenchymal stem cells using minicircles with the same four reprogramming factors, but without the need for a layer of feeder cells supplying additional nutrients and reprogramming factors [147]. Fernandes and Chari engineered neural stem cells to secrete brain-derived neurotrophic factor (BDNF) using magnetic nanoparticles in conjunction with minicircle or plasmid DNA [148]. The minicircle had over five-fold higher transfection efficiency than the plasmid, and minicircle-transfected cells had a ten-fold increase in BDNF secretion over plasmid-transfected cells [148]. These pre-clinical tests demonstrate that stem cell reprogramming is feasible with minimized DNA vectors.

Therapy for the Lungs
Lungs are easily accessible via aerosolization [149,150]. Thus far, however, gene therapy approaches for pulmonary diseases, including cystic fibrosis, cancer, and asthma, have not been fully realized. Impenetrability of the mucosal layer in the lung to gene therapy vectors (viral or non-viral), and only limited and transient gene expression explain the difficulties encountered in prior attempts [151][152][153]. Minimized vector systems have the potential to deliver to lungs because naked DNA minivectors smaller than 1,200 bp can be aerosolized [71].
Minicircles have been effectively used both in vitro and in vivo to target lung epithelial cells with enhanced GFP (eGFP), firefly luciferase (Luc), or DNAH5, which encodes an outer dynein arm protein involved in primary ciliary dyskinesia [150]. Minicircles carrying these genes displayed higher levels of gene expression compared to plasmids [150]. Higher and more prolonged gene expression afforded by minicircles or minivectors reduces the number of therapeutic treatments needed. Fewer vector administrations reduces the potential for adaptive immune responses, such as are seen with the use of viral vectors or plasmids (reviewed by [154]). In this regard, reduction of CpGs in plasmid vectors or the use of CpG-free plasmids permits repeated delivery to lungs without causing toxicity or immunogenicity [155,156]. Minimized vectors such as minivectors do not contain CpG motifs (unless they are specifically added as part of the design) [73]. Furthermore, the nanosize (~5 nm × 5 nm × 45 nm) of supercoiled minivectors may prevent them from getting stuck in the mucosal layer, allowing them to penetrate cells and improve lung therapy outcomes. Together, these findings not only mean that minivectors could be used to deliver genes to address lung-specific diseases, but may also enable the lung to be used as a route into the body for the systemic delivery of therapeutic molecules into the bloodstream.

Cardiovascular Uses
One of the most significant health issues facing the world today is cardiovascular disease. The enormous potential as well as the particular challenges of using gene therapy for cardiovascular diseases were recently reviewed [157]. The pitfall of reduced gene expression of naked plasmid injections (a standard delivery technique for gene therapy of the heart) was addressed by comparing delivery of minicircle and plasmid encoding the same target [158]. Minicircle expressed similarly to plasmids [158], thereby constituting a safer alternative to plasmid vectors for treatment of heart diseases. Minicircles may be the better alternative to viral vectors because of viral-associated immunogenicity issues, as reported in a study aimed to treat familial hypercholesterolemia [159]. Finally, minicircles encoding either shRNA or miRNA also renew hope of combating heart disease with gene therapy [160,161], particularly because of their size advantage compared to other vectors, and especially when combined with short therapeutic sequences.

Dermal Uses
The skin is a readily accessible target for non-viral gene therapy. One application for gene therapy of the skin is wound healing in diabetics, specifically by enhancing expression of vascular endothelial growth factor (VEGF). VEGF triggers the growth of new blood vessels, bringing more healing factors to a wound. Yoon et al. enhanced wound healing in diabetic mice by subcutaneously injecting vectors expressing VEGF in a microbubble solution, then forcing transfection by popping the bubbles with ultrasound [162]. They found that a minicircle construct expressed VEGF more effectively than the plasmid constructs they tested [162].
DNA vector treatments can be applied for psoriasis, an autoimmune skin condition caused by chronic inflammation and dysregulated angiogenesis [163]. Expressing a gene that inhibits transforming growth factor-β prevents aberrant angiogenesis that can treat not only psoriasis [163] but also improve wound healing in the cornea [164]. The most logical method for delivering gene therapy vectors to the skin is topical application of a vector, which is theoretically possible. Subcutaneous injections would improve access of the vector to dermal immune cells. Using sonoporation to precisely release DNA vectors from microbubbles only at the wound or plaque psoriasis site prevents distant cells from being exposed to unwanted immune-stimulating or immune-suppressing genes.

Concluding Remarks: Moving Minimized Non-Viral DNA Vectors into the Clinic
Worldwide, the total number of reported active gene therapy clinical trials is 2,400 [165]. Non-viral DNA vectors-plasmids, oligonucleotides, and the pCOR, pORT and MIDGE systems-have been thus far used less frequently than viral vectors (21% vs 79%). Other non-viral vector systems described in this review, such as pFAR and minicircles, have been tested only pre-clinically. Some of these pre-clinical studies have been described in the reviews of Vandermeulen et al. [45], Gaspar et al. [9], and Wong et al. [115]. In this review, we addressed some of the specific minicircle studies that occurred since those reviewed by . Table 1 provides the status of non-viral DNA vectors in clinical trials. So far, only pCOR, pORT, and MIDGE are listed in this table, but recent advancements made in minimized DNA vectors should increase these numbers. Numerous successful pre-clinical studies will lead to increased numbers of minimized DNA vectors in clinical trials for some of the most sought-after targets in gene therapy, like cancer and HIV. Efforts continue to make production of minimized DNA vectors more cost effective and to improve their purity [92][93][94][95].
Ideal for delivering short hairpin RNA [72,166], miRNA [160], cytokines [139], and other biologics, such as synthetic protein drugs [167,168], minimized DNA vectors allow for multiple important uses. Minicircles have even been co-delivered with drugs [169][170][171]. Additional future applications of minimized DNA vectors include engineering them to recognize specific cells or organs, construction of genetic logic gates to control gene expression [172,173], and the potential to control the shape of specific vectors [174]. To maintain the advantages of small vector size, delivery of long genes could be achieved by breaking such genes into multiple fragments and encoding each fragment on multiple vectors [172,175]. Gene fragmentation and protein reconstitution [176] are tools that can be easily incorporated into minimized DNA vectors. Furthermore, minimized DNA vectors are potentially better suited for the treatment of polygenic diseases because multiple vectors can be delivered simultaneously and administered repeatedly as needed.
In this review, we have highlighted numerous ways researchers have endeavored to realize the potential of gene therapy using non-viral DNA vectors. It is critically important that more work be done to understand the advantages and disadvantages of each tool in the gene therapy toolbox. Most of the advances made to non-viral DNA vector engineering have involved reducing length and removing problematic DNA sequences. Each of these advances, however, brought new problems to solve-problems of purity and scale. These new issues are only now beginning to be addressed; nevertheless, the steady progress and successes are encouraging. The improvements afforded by decreased cellular toxicity, increased transfection efficiency, increased number (moles) of payload delivered per mass, enhanced purity and yield, and decreased expense all combine to make non-viral DNA vectors a favorable gene therapy tool.

Acknowledgments:
The authors were supported in part by the Baylor-University of Texas (UT) Houston Center for acquired immune deficiency syndrome (AIDS) Research Core Support Grant number P30 AI036211 (Butel, PI), and in part by the National Institutes of Health (NIH) grants R56 AI054830 and R01 GM115501 to L.Z. We thank our generous anonymous donors for the Research Support Fund. We thank Dr. Brian E. Gilbert (Baylor College of Medicine, Houston, TX, USA) and Christopher E. Coker (Twister Biotech, Inc., Houston, TX, USA) for critically reading the manuscript.
Author Contributions: All authors contributed equally to the drafting, editing, and proofreading of this review.

Conflicts of Interest:
The funding sponsors had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.
Lirio M. Arévalo-Soliz and Lynn Zechiedrich are co-inventors on issued and pending patents covering the minivector technology in this paper.