G. Haegeman, Department of Molecular Biology, University of Gent-VIB, K.L. Ledeganckstraat 35, B-9000 Gent, Belgium, Fax: + 32 9264 53 04, Tel.: + 32 9264 51 66, E-mail: Guy.Haegeman@dmb.rug.ac.be
An important reason for preferring mammalian cells for heterologous gene expression is their ability to make authentic proteins containing post-translational modifications similar to those of the native protein. The development of expression systems for mammalian cells has been ongoing for several years, resulting in a wide variety of effective expression vectors. The aim of this review is to highlight episomal expression vectors. Such episomal plasmids are usually based on sequences from DNA viruses, such as BK virus, bovine papilloma virus 1 and Epstein–Barr virus. In this review we will mainly focus on the improvements made towards the usefulness of these systems for gene expression studies and gene therapy.
The development of eukaryotic expression vectors has provided a direct and convenient way of introducing novel genetic information into cultured cells. Heterologous expression of recombinant proteins has subsequently found widespread use for overproduction of therapeutically important proteins and for studies of gene regulation [1–7]. However, integration of foreign DNA into the genome of established cells and organisms may have important consequences, especially when used for gene therapy. DNA insertions may have far-reaching effects such as insertional mutagenesis or silencing by de novo methylation of the introduced gene or nearby cellular DNA [8–10]. To circumvent problems inherent in genomic integration, the genetic engineering of episomal (extrachromosomal) eukaryotic vectors offers an attractive alternative.
One of the first virus-derived vectors for introducing genes into cells was based on genetic elements of the polyomavirus simian virus 40 (SV40). SV40 exhibits a replication pattern that is uncoupled from the regulatory mechanisms of the host cell, so that each viral genome replicates many times within each cell cycle. As a consequence, transfection of permissive cells with recombinant SV40 vectors results in cell death, limiting this vector to transient expression set-ups. For stable expression experiments, episomal vectors were designed on the basis of viral elements of BK virus (BKV), bovine papilloma virus 1 (BPV-1) and Epstein–Barr virus (EBV). Each of these vectors contains a viral origin of DNA replication and a viral early gene(s), the product of which activates the viral origin and thus allows the episome to reside in the transfected host cell line in a well-controlled manner. There are several advantages of episomal vectors; first, the inserted gene of interest cannot be interrupted or subjected to regulatory constraints which often occur from integration into cellular DNA. Secondly, the presence of the inserted heterologous gene does not lead to rearrangement or interruption of the cell’s own important regions. Thirdly, episomal vectors persist in multiple copies in the nucleus, resulting in amplification of the gene of interest, whenever the necessary viral trans-acting factors are supplied. Finally, in stable transfection procedures, the use of episomal vectors often results in a higher transfection efficiency than the use of chromosome-integrating plasmids [11,12]. Furthermore, episomal vectors are plasmid constructions that replicate in both eukaryotic and prokaryotic cells and can therefore easily be ‘shuttled’ from one cell system to another. This feature and the fact that some episomal vectors can transfer large amounts of DNA allow applications, such as screening of cDNA libraries [13–15]. Because episomal vectors utilize the cellular enzymes for replication and repair, they are also powerful tools for studying DNA replication or mutagenesis [16,17].
Recently, episomal vectors have also been considered for use in gene therapy [18,19]. Oncogenes are often expressed at high levels in malignant cells. One approach to gene therapy of human cancer cells is to use antisense strategies to block expression of dominant oncogenes. Another interesting application of gene therapy is the delivery of genes to cells in order to complement a default gene causing a severe disease [11,20]. However, high-level expression of these genes presents a considerable technical obstacle . One method of achieving high expression levels without affecting the cellular genome is to use episomal plasmid vectors, which replicate extrachromosomally.
Historically, the polyomaviruses were grouped together with the papillomaviruses to form the papovavirus family ( Table 1). Members of the polyomavirus and the papillomavirus subfamilies are readily distinguished by differences in the size of the virion (40 nm vs. 55 nm, respectively) and by the size of the viral genomes (5000 bp vs. 8000 bp) . Sequence comparison indicates that the genomic organization of the two subfamilies is distinct, suggesting that these two groups of viruses are evolutionarily unrelated . BKV is a human polyomavirus ( Table 1), and was first isolated in 1971 from the urine of a renal transplant recipient . It is ubiquitous in the human population all over the world, and primary infection occurs in childhood. BKV antibodies are detected in 50% of 3-year-old children, and almost all individuals have been infected by the age of 10. Primary infections with BKV are only occasionally associated with clinical conditions and are usually followed by a persistent latent infection which may be re-activated under conditions of impaired immunosurveillance. The main site of BKV latency in healthy people is the kidney. Immunosuppression later in life may lead to viral re-activation and foci of infection in the kidney epithelium. BKV may cause upper respiratory or urinary tract disease [25–27].
Table 1. Classification and characteristics of viruses.
Genome size (bp)
Natural infection of Asian macaques; persists in kidney
Common infection in early childhood; persists in kidney and in lymphocytes
Cutaneous fibropapillomas; can transform rodent cells in vitro
Common infection in early childhood; persists latently in lymphocytes
All members of the Polyomavirinea display a similar genomic organization, and the DNA sequence homology between, e.g. BKV and the monkey polyomavirus SV40, is greater than 80% [28–31]. The virion is a 40- to 45-nm icosahedral particle containing a circular, double-stranded DNA molecule of 5153 nucleotide pairs  with a chromatin-like nucleoprotein structure. The genome is functionally divided into an early region (2.4 kb), which codes for the large T and the small t antigen, a late region (2.3 kb), which codes for the viral capsid proteins, VP1, VP2 and VP3, and a noncoding regulatory region (0.4 kb). The early region is, by definition, the portion of the genome transcribed and expressed early after the virus enters the cells, i.e. before the onset of viral DNA replication. The late region of the genome is only efficiently expressed after viral DNA replication has been initiated. Early and late transcription both start near the origin of replication (ori) and proceed in opposite directions on separate DNA strands. The regulatory region of the viral DNA is located between these early and late regions and contains the large T antigen-binding sites, a unique origin of DNA replication, and an early enhancer region consisting of three repeats and an element called c ( Fig. 1A) [32–35]. The three repeats act as transcriptional enhancers, but are also necessary for DNA replication . A schematic representation of BKV large T antigen is given in Fig. 1B. The N-terminal region of BKV large T antigen, starting from the initiation codon up to the nuclear localization signal, plays an essential role in viral replication, transactivation and transformation. The middle part is also important for replication and contains regions for binding of T antigen to BKV DNA and a zinc finger region. The helicase and ATPase activity are contained in this region, as well as sequences important for complex formation with the cellular p53 protein. The C-terminal part is most divergent between the sequences of different polyomaviruses. A helper function, which enables the virus to grow efficiently, has been suggested for this domain. This domain has also been shown to display an activity, termed host range function, that is required late in viral productive infection, probably in viral assembly. This C-terminus is dispensable for viral DNA replication . All polyomaviruses multiply in the nucleus. Viral DNA is in the form of a minichromosome, and replication occurs bidirectionally, starting with the binding of T antigen to binding site II in the origin region ( Fig. 1A). These interactions are complex and much remains to be learned about the protein–DNA interactions that take place at the ori, the structural consequences of these interactions, and the way these processes are regulated. This has been studied in much more detail for SV40 [38–40]. Considering the high degree of homology between the two viruses, it is likely that a similar mechanism is used by BKV. In the presence of ATP, T antigen undergoes a conformational change, permitting the assembly of a bilobed double hexamer of T antigen at the origin of DNA replication. A variety of physical and chemical analyses indicates that the bound T antigen double hexamer catalyzes the local unwinding of part of the early palindrome and distortion of the A/T elements. Once this initial unwinding has occurred, T antigen associates with replication protein A (RPA), forming a preinitiation complex. RPA is a single-stranded DNA-binding protein that keeps unwound regions single-stranded, which allows more extensive unwinding of the DNA as a result of the helicase activity of the T antigen. Recruitment of DNA polymerase α-primase to this preinitiation complex converts it to a functional initiation complex. This recruitment step in viral DNA replication defines the very limited host range for members of the Polyomavirinae.
Bkv sequences as elements in gene expression vectors
BKV efficiently infects human cells and transforms hamster, mouse, rat, rabbit and monkey cells in tissue culture. Infected human cells harbor viral DNA mostly as unintegrated episomal molecules. This interesting feature forms an attractive starting point for the development of BKV-derived gene expression vectors. Stable replication of BKV-derived vectors within transfected cells requires a functional ori, large T antigen with its intact DNA-binding and helicase activities, and the set of cellular proteins involved in replicative DNA synthesis. As BKV DNA is infectious for human cells, the late viral sequences are deleted in most constructed BKV-derived vectors and replaced by heterologous coding sequences.
Figure 1C represents a typical BKV-derived vector for stable gene expression in mammalian cells. The viral part of the vector consists of the BKV early regulatory region, i.e. the BKV ori and the BKV T antigen. These sequences are necessary for stable maintenance of the vector in the cell. A eukaryotic selection cassette, such as the neomycin resistance gene and the adjacent regulatory regions, makes it possible to screen for transfected cells. The unusual feature of BKV to stably replicate as an episome in certain cell types was first demonstrated by Milanesi and collaborators . The BKV-derived vectors described include almost the entire BKV genome, except for the sequences of the late viral capsid transcripts, the thymidine kinase gene of herpes virus simplex type 1, and bacterial plasmid sequences [41–45]. These vectors have already been successfully used for the efficient expression of different proteins including several reporter gene proteins [43,46–50], the herpes simplex virus type 1 glycoprotein B [41,44,45], and the human serotonin receptor h5-HT1B. BKV-derived vectors have also been reported to be useful tools for gene delivery into less convenient cell types, such as the human bladder carcinoma cell line HT-1376 .
BKV episomes appear to replicate once per cell cycle in stable transfectants. In the majority of independently transfected human cells, BKV-derived vectors persist in the episomal state. Removal of selective pressure did not affect the episomal status of the BKV-derived molecules, indicating that the copy level is probably held constant by autoregulation of BKV T antigen synthesis . The copy number of the BKV-derived vectors depends on the nature of the human cell line used, but also on the composition of the expression vector itself. Levels for BKV-derived vectors varied from 20 to 40 copies in HeLaH21 and 293HEK cells to 75–120 copies in 143B cells. Correspondingly, the inserted cDNA is amplified, which results in a higher expression of the protein of interest. The quantity produced correlates well with vector copy number, suggesting that the critical factor for abundant expression is amplification of the cDNA [46,47]. Only negligible amounts, if any, of this vector DNA integrate into chromosomal DNA.
Bovine papilloma virus 1
The papillomaviruses are widespread in nature and have been recognized primarily in higher vertebrates. They are small double-stranded DNA viruses that cause warts in the epithelial tissues covering many different mammalian organs. In most cases, these lesions are self-limiting, but occasionally they slowly progress to malignancy . Certain papillomaviruses, including the bovine papillomaviruses, are capable of transforming rodent cells in vitro, but virus production has never been successful in tissue culture. Of the papillomaviruses, BPV-1 is probably the most intensively studied. It consists of a single molecule of double-stranded circular DNA, contained within a spherical protein coat, or capsid, composed of 72 capsomeres. The capsid consists of two structural proteins L1 and L2. The viral genome, which is associated with cellular histones to form a chromatin-like complex, can be functionally divided into two domains: the E (early) region, which contains eight ORFs, encoding proteins involved in DNA replication, regulation of transcription and cellular transformation, and the L (late) region, containing two large ORFs coding for the structural polypeptides of the virion, L1 and L2. All the ORFs are located on the same strand of the viral DNA [22,23].
In virally transformed rodent cells, plasmid DNA replication can be divided into two phases: establishment and maintenance . There is an initial establishment period during which the viral genome is amplified from a low copy number to a final, moderate copy number of ≈ 20–300 copies per cell. Although BPV-1-derived plasmids are not subjected to once-per-cell cycle replication [53–55], they are stably maintained during the subsequent maintenance phase. The level of an essential factor, which is kept constant, may determine the final copy number. In this model, amplification would occur as long as the factor is in excess relative to the copy number, while maintenance would be obtained when the factor becomes limiting. E2, which seems to be the master regulator of papillomavirus replication and transcription, is one candidate for this role as limiting factor [56,57]. BPV-1 DNA replication requires the ori in cis and the viral E1 and E2 proteins in trans[52,58,59]. Although E1 appears to be the factor that is directly involved in DNA replication, an auxiliary role of E2 is required to alter the DNA-binding activity of E1. E1 displays only a limited sequence specificity, but the interaction with E2 during the initiation of replication results in a co-operative, highly sequence-specific binding on the BPV-1 ori [60–64]. The assembly of E1 at the origin is followed by distortion of the DNA at the A/T-rich sequence. The helicase activity of E1 then initiates unwinding of the DNA, followed by recruitment of RPA [65,66]. E1 also interacts with cellular DNA polymerase α-primase . Cellular replication proteins are recruited to the origin and an active replication complex is formed . E2 may relieve nucleosome-mediated repression of papillomavirus DNA replication in vitro and may stimulate DNA replication by recruiting host replication factors to the origin, such as the host cell single-stranded DNA-binding protein RPA . The minimal ori ( Fig. 2A) is absolutely required, but is not sufficient for stable extrachromosomal replication of the viral plasmid. An additional element, the minichromosome maintenance element, in the upstream regulatory region of BPV-1 ( Fig. 2A) ensures stable replication and is also important in long-term episomal maintenance of BPV-1 in cells expressing E1 and E2 . Long-term episomal maintenance of BPV-1 plasmids in the cell is mediated through a noncovalent association of the viral genomes with cellular chromosomes during mitosis. Analysis of viral factors revealed that the E2 protein in trans and its multiple binding sites in cis in the minichromosome maintenance element are both necessary and sufficient for the chromatin attachment of the plasmids. The segregation activity of E2 is separate from its enhancer role in gene expression or replication and appears to be regulated by phosphorylation [69–71]. Mutation studies suggest that E1 may also be a critical component in the segregation mechanism . The E1 ORF ( Fig. 2B) is relatively well conserved among all the papillomaviruses and even shares structural similarities with large T antigen of SV40. These similarities include regions of ATPase, helicase, and DNA-binding activity [65,66,72].
Bpv-1-derived vectors for heterologous gene expression
A subfragment of BPV-1 representing 69% of the viral DNA, consisting mainly of the E region, is sufficient to transform mouse cells, while it may still be maintained as a stable extrachromosomal multicopy element without being infectious [73,74] ( Fig. 2C). The use of this fragment in a cloning vector resulted in efficient production of rat preproinsulin and human growth hormone in mouse C127 cells [74,75]. In these early studies, transformants were picked up on the basis of their altered morphology. As this selection criterion is not always evident and several cell types are not sensitive to morphological transformation by BPV-1, the same BPV-1 fragment was combined with a eukaryotic resistance gene in later studies . Such BPV-1-based vectors have been successfully used to generate cell lines that synthesize a number of secretory and membrane proteins, such as human growth hormone , influenza virus hemagglutinin , and vesicular stomatitis virus glycoprotein G .
Vectors containing the entire BPV-1 genome  or the E region are in many cases propagated as stable multicopy extrachromosomal elements in transformed cells. The copy number ranges from 20 to 300 copies per cell. An explanation for this extensive cell to cell variation has not yet been found. Although extrachromosomal BPV-1 molecules replicate independently of the chromosomal DNA in their host, the copy number of the viral plasmids remains relatively constant for many cell generations . Despite the fact that BPV-1-derived vectors have proven their usefulness in heterologous expression experiments, they never found a widespread use. This is mainly because they are only maintained as plasmids in a limited number of host cells including those of mouse and rat origin. Most expression studies have been performed in the mouse cell line C127. Furthermore, it has also been reported that these vectors may show a high frequency of DNA rearrangements  and that they may integrate into chromosomal DNA as head-to-tail tandem arrays [77,79,81]. This indicates that the copy number and physical state of the vector DNA appears to be dependent on both its composition and the nature of the host cell. Further improvements came from the development of a vector without detectable transforming activity. This vector, which contains only E1, E2 and the upstream regulatory region of BPV-1, is stably maintained in several mouse and human cell lines, such as HeLa; the expression of interleukin 2 and CD80 was for most cells similar to that of the parental vector (69% of the BPV-1 vector) . This vector may be useful and safe for human gene therapy, as will be discussed below.
Episomal vectors based on EBV components have found significant and increasing utility in biotechnology . EBV is a member of the Gammaherpesvirinae ( Table 1) which establishes latent infection in lymphocytes and is associated with cell proliferation. About 90% of the world’s adult population is estimated to be infected with EBV. Most people become infected with the virus in early childhood and carry the virus latent over a long period of time. If the initial infection is delayed until adolescence, a variety of neurological syndromes may occur, clinically manifested as infectious mononucleosis. Much of the interest in the gammaherpesviruses comes from their association with cell proliferation and cancer. EBV may directly cause B lymphoproliferative disease in immune-deficient humans and is an etiologic agent in Burkitt’s lymphoma, Hodgkin’s disease, unusual T-cell lymphomas and nasopharyngeal carcinoma [83–86]. Besides infectious mononucleosis, EBV may also cause a variety of other neurological syndromes. The capability of infecting astrocyte cell lines suggests that the virus may also play a role in the pathology of EBV in the brain .
A typical herpes virion consists of a core, containing a linear double-stranded DNA of 172 kb , an icosadeltahedral capsid ≈ 100–110 nm in diameter with 162 capsomeres, a protein tegument between the nucleocapsid and the envelope, and an outer envelope with glycoprotein spikes on its surface [84,89,90]. During latency, about 10 of the ≈ 100 genes encoded by the virus are differentially expressed, depending on the infected host cell. They comprise six Epstein–Barr nuclear antigens (EBNA1, 2, 3A, 3B, 3C and LP) and three integral membrane proteins (LMP1, 2A and 2B). In addition, two genes coding for small RNAs (EBER1 and EBER2) are actively transcribed in most latently EBV-infected cells [85,91]. During the lytic cycle, however, most EBV genes are expressed. In the latent cycle, EBV replication is mediated by oriP, a cis-acting element identified as the origin of DNA replication [92–94], while during the lytic phase, ori Lyt is activated. The viral life cycle of EBV has been reviewed by Kieff .
In this review, we will focus on two sequences, oriP and EBNA1, the only viral elements necessary and sufficient for stable episomal maintenance of viral DNA in the cell. As stable episomal maintenance reflects the sum of plasmid replication and efficient segregation to the daughter cells, both aspects will be described in more detail.
EBV vectors, but also plasmids containing only oriP, replicate once per cell cycle in synchrony with the host chromosomes [95,96]. It has been shown that, in latently infected cells, EBV vectors, which themselves possess a chromatin-like structure , are anchored to the nuclear matrix through a high-affinity matrix attachment region, containing oriP [98,99]. OriP is composed of two noncontiguous regions: the family of repeats (FR) and the dyad symmetry (DS) element ( Fig. 3A). The FR consists of 20 tandem imperfect copies of a 30-bp repeat sequence, while the other region, the DS, contains four related copies of the 30-bp repeat. These 30-bp repeats contain consensus sequences for EBNA1 binding [94,100]. EBNA1 is a direct DNA-binding protein and has a helix–turn–helix DNA-binding moiety [101,102]. A structural distortion in the DS upon binding of EBNA1 dimers is thought to be important for the initiation of DNA synthesis [103–105], which occurs near the DS element. Replication requires EBNA1 in trans; the actual contribution of EBNA1 to replication from oriP is still a matter of debate, and no enzymatic activities correlated with replication, such as ATPase or helicase function, were found to be intrinsic to EBNA1 . As EBNA1 is the only virally encoded protein necessary for plasmid maintenance, the host cell replication apparatus mainly performs replication. Gel-shift experiments with HeLa cell nuclear extracts revealed specific binding of cellular factors to oriP. These proteins compete with EBNA1 for binding to oriP in vitro, suggesting that these cellular factors regulate EBV replication by displacing EBNA1 at the origin [107–109]. Replication initiation is greatly stimulated by the FR, possibly by formation of a DNA loop between the FR and the DS sites, brought about by the interaction of EBNA1. This association appears to increase the apparent affinity of EBNA1 for DS and may contribute to initiation of replication at oriP . DNA looping is a mechanism by which enhancer or repressor elements act at a distance to activate or repress promoter elements [111–113]. Such interaction may be important to affect DNA structure to allow binding or attraction of replication proteins to initiate DNA replication [114–116]. Replication is initially bidirectional, but the FR contains a replication fork barrier and therefore the replication from oriP proceeds mainly unidirectionally [117–119]. This controlled replication allows the production of stable vectors that do not have an elevated mutation rate  and do not readily rearrange .
Recently, it was shown that the latent ori of EBV is more complex than formerly appreciated: it is a multicomponent origin of which the DS element is one efficient component for replication, but in which other regions do also contribute to replication. Flanking the DS is an element called Rep*, which may partially substitute for the DS in the support of replication. Besides this Rep* element of the EBV genome, other eukaryotic sequences that mediate DNA synthesis also support long-term replication . As an alternative to the whole oriP region, only the FR on its own may be used if another replication origin is provided, for example by including human genomic DNA. Krysan et al.  made use of such a defective EBV-derived vector to screen different human genomic sequences which replicate autonomously when introduced into human cells. Therefore, the DS element was replaced by different genomic sequences and the isolated sequences represented authentic human origins of replication [120,121].
Recent data indicate that, during latent infection in vivo in proliferating cells, the total EBV chromosome does not require the replication-initiation function of oriP. The essential role of EBNA1 would only be the maintenance of the circular EBV chromosome, almost certainly in conjunction with the extrachromosomal maintenance function of oriP [122,123]. The role of EBNA1 in replication of artificial oriP-containing plasmids is still a matter of debate. Aiyar et al.  demonstrated that EBNA1 is not required for the synthesis of oriP plasmids. This is in contrast with recent results of Yates et al.  who concluded that the DS of oriP is an EBNA1-dependent replicator. It has been clearly demonstrated that EBNA1 bound to oriP functions postsynthetically to ensure plasmid maintenance and segregation in dividing cells. This retention mechanism requires EBNA1 binding to the FR on the plasmid and to chromosomal components, thereby mediating physical association of the plasmid with the chromosome [110,126]. With the use of yeast two-hybrid technology, a cellular protein partner for EBNA1, EBP2 (EBNA1-binding protein 2), was found which seems to be important for stable segregation of EBV episomes during cell division, but not for replication of the episomes .
A third feature of oriP in the presence of EBNA1 is a transactivation function. The FR of oriP may, on binding of EBNA1, enhance replication and retention, as well as transcription . This transcription-enhancer function, which involves at least six or seven 30-bp repeats for full activation , is dependent on the cell line used and on the promoter driving the gene of interest . How EBNA1 mediates this transactivation function can only be hypothesized. Data of Wang et al.  suggest that P32/TAP, which interacts strongly with EBNA1, may contribute to EBNA1-mediated transactivation. P32/TAP has been implicated in splicing, transactivation, and receptor functions. Besides the fact that the nuclear localization domain improves the transactivation function of EBNA1, it has also been shown that DNA linking by EBNA1, through its three linking domains ( Fig. 3B), contributes to the activation of transcription and replication [131,132].
Ebv as a gene expression system
A standard EBV-derived, episomal vector used for gene expression carries the origin of replication (oriP) and a sequence encoding a trans-acting factor EBNA1 ( Fig. 3C). As stated above, these two sequences are necessary and sufficient for retention and replication of the EBV vector in a variety of uninfected established cells, including human epithelial, fibroblast and lymphoma cells, as well as monkey and dog cell lines [59,92,93,133,134]. A commonly used EBV expression plasmid, p220.2, also carries the gene conferring resistance to hygromycin for selection in mammalian cells and a fragment of pBR322 for selection (ampicillin) and replication in bacteria ( Fig. 3C). The eukaryotic resistance gene is necessary because the retention of EBV-derived plasmids is imperfect and results in slow loss of the vector over time. Retention rates of 92–98% per cell generation are typical of EBV vectors in the absence of selection, which is still far superior to the high loss of plasmids lacking replication and retention elements [82,93,135,136]. Plasmid maintenance reflects the sum of plasmid replication and segregation. Fluorescence in situ hybridization data provide strong evidence of association of the EBV-derived vectors with host cell metaphase chromosomes. This observation indicates a variation in episomal copy number between metaphase spreads within each cell line, suggesting that the partitioning of the vectors is not exact. However, this passive method of episomal segregation, while not matching the strictly controlled segregation of endogenous eukaryotic chromosomes, does appear to be effective, allowing the persistence of molecules over prolonged periods of culture [126,137]. There is a considerable variation of viral gene copy number among different cell lines , ranging from 5 to 100. During prolonged propagation, little variation in EBV copy number has been observed. The persistence of multiple copies results in amplification of the gene of interest and higher protein expression in a relatively short period of time. Even in a transient experiment, EBV vectors give higher expression levels of the gene of interest in various cell types [12,94,128]. EBV-derived expression vectors have already been successfully applied in several studies, including expression of cytokine genes [136,139], cytokine receptors , growth factors , amino-acid transporter , full-length human α1 chain of collagen V , etc. Also less convenient cell types, such as human peripheral blood CD34+ cells, the peripheral neuroepithelioma cell line CHP-100, the neuroblastoma cell line IMR-32, and human bone marrow cells could be successfully transfected in vitro with an EBV-based vector system [4,144,145]. EBV-based vectors were also used for antisense RNA-mediated inhibition of gene expression, e.g. inhibition of protein kinase Cα expression by antisense RNA in transfected Jurkat cells . EBV-derived vectors without EBNA1 gene may be used successfully, if the trans-acting factor EBNA1 is supplied from a constitutive expression plasmid [142,143,147–150]. Horlick et al. demonstrated that this system rapidly generates stable cell lines suitable for scale-up . EBV-derived vectors are in most cases stably maintained as episomal vectors in the cell. However, in some cases integration of the vector in the host chromosome has been reported . Also rearrangements of the EBV-derived vectors have been observed in the cell line A431 . These reports indicate that the behavior of the EBV-derived vectors depends on the cell line.
EBV-based vectors are mainly used in primate cells. Early reports mentioned the failure of EBV-derived vectors to replicate in mouse and hamster cells [93,129]. This statement forms a serious drawback for gene therapy approaches (see further) because testing EBV vectors in animals, such as mice, would not be possible. Further investigations were performed and, although rodent cells are not permissive for EBV, it was elucidated that, when these cells were transiently transfected with an EBV vector, they expressed the marker gene more intensively than with a conventional plasmid vector . Furthermore, EBNA1 is adequately expressed from p220.2 in hamster cells, whereby EBNA1 and the FR appear to supply the same nuclear retention function in hamster cells as they do in human cells . The failure of replication in mouse and hamster cells was mainly ascribed to oriP; therefore, in stable transfection set-ups, other sequences were inserted into the EBV vector, e.g. replacement of a part of oriP (DS) with sequences that mediate efficient replication in that organism [155–157]. Nevertheless, some authors discussed the stable replication of ‘original’ oriP-containing vectors in some types of rodent cells [158–160]. In these experiments, the EBV-derived expression plasmid contained no additional sequences for stabilization of the EBV vector in rat cells, and experiments clearly showed that the EBV vector has replicated in PC12, L, L6 and C6 cells.
EBV-derived vectors are also often used as shuttle vectors. Their ability to replicate in both eukaryotic and bacterial systems has made them useful tools in mutation studies and for screening of libraries. After transfection of a mammalian cell line, the episomal DNA can easily be isolated [161,162] and used to transform bacterial cells such as Escherichia coli. The first shuttle vectors were based on SV40 sequences, but these experiments revealed a high spontaneous mutation frequency. Also, SV40 vectors replicate poorly in most human cells. Vector systems based on the EBV viral replication and retention components may be more useful in several research areas . The low spontaneous mutation frequency of EBV, i.e. less than 105, makes these vectors useful vehicles in mutation studies. The EBV replicon and retention system is capable of carrying 172 kb of viral DNA ; this may be useful when screening genomic libraries [13,14,163] or for detecting replication sequences, as discussed previously [120,121].
Human artificial episomal chromosomes based on the latent replication origin of EBV have been developed for the propagation and stable maintenance of DNA circular minichromosomes in human cells . This system provides an important tool for the functional study of large mammalian DNA regions and gene therapy. Many functional units in human cells span hundreds of kilobases. In addition, regions far from the transcription unit may be critical for proper expression. In order to study gene function and regulation, it would be interesting to transfer an entire functional unit as a single DNA fragment into human cells. Large-insert cloning systems, such as yeast or bacterial artificial chromosomes, do not always appear suitable for the transfer and functional analysis of large DNA fragments in human cells. The DNA often shows a high degree of internal rearrangement or deletion and is therefore not stably maintained, unless it becomes integrated into mammalian chromosomes . To overcome this problem, the oriP has been incorporated into the yeast artificial chromosome system . As this technique requires large amounts of DNA, which is difficult to purify from yeast, vectors based on the bacterial artificial chromosome vector in combination with oriP were constructed [167,168]. Such systems allow the analysis of large human genes or gene clusters in their natural configuration and genetic background, study of the influence of remote control elements on gene expression, and analysis of long-range effects of regulatory elements, such as those involved in chromosome inactivation and imprinting.
Episomal virus-derived vectors ingene therapy
In theory, gene therapy seems an easy technique. The delivery of corrective genetic material into cells is meant to alleviate the symptoms of the disease or even to correct a defective gene. In practice, considerable problems have emerged, as reviewed by Verma . Optimal vectors for gene therapy require (a) high-level and stable expression of the gene of interest, (b) a high transfection efficiency, (c) no integration into the chromosomal DNA to avoid effects on the cell’s own DNA and on the vector itself, and (d) no transformation features that may result in secondary cancers. Vectors meeting all these criteria are not available [21,81,169,170]. Gene transfer by nonviral vector-mediated systems has been shown to be a safe and simple but relatively inefficient method for gene delivery in vivo. Many of the systems currently available utilize viral vectors derived from retroviruses, lentiviruses, adenoviruses, etc. [21,171]. We will focus on the utility of nonintegrating viruses for the construction of an efficient gene-delivery vector. Expression vectors derived from BKV , BPV-1  and EBV [18,172–174] have been explored for the transfer of a foreign gene to a target organ. The most difficult condition the vector has to fulfil is efficient gene delivery. Especially when episomal vectors are carrying multiple genes or native versions of complete genes with introns and endogenous control regions, the vector size becomes problematic. Successful transfection procedures for these vectors include electroporation , but also combinations of the plasmids with cationic liposomes [12,175] or with liposome particles derived from ‘hemagglutinating virus of Japan’[144,176,177]. Combination of an episomally (BKV) replicative DNA plasmid with a liposomal delivery system resulted in long-term expression (up to 3 months) of transgene luciferase in mice .
Until now, most progression towards the development of an efficient episomal gene therapy vector has come from EBV vectors. Intramuscular injection of EBV-based dystrophin expression plasmids into mice resulted in a significant enhancement in number of muscle fibres expressing recombinant dystrophin compared with experiments with a conventional vector. Such a vector may have its therapeutic applications in Duchenne muscular dystrophy . In human cells without selection medium, episomal EBV vectors have a prolonged but not an indefinite retention. Therefore, when the goal is to quickly kill tumor cells, the use of EBV vectors may be a good strategy. EBV vectors have been tested in human B-cell lymphomas, as a model for immune therapy of cancer. The vectors were able to stably and efficiently express cytokine genes over a period of several weeks, in tissue culture and after injection of the transfected cells into nude mice [136,173]. EBV-based vectors are also successfully used for the correction of deficient human cell lines by cDNA or genomic DNA transfections leading to complementation and cloning of the correcting gene [11,20]. The use of vectors carrying only oriP has been suggested as a strategy to selectively target tumor cells expressing EBNA-1, as is the case in EBV-associated neoplasms, such as Burkitt’s lymphoma and nasopharyngeal carcinoma [174,176].
The major drawback of such episomal expression vectors in gene therapy is the fact that they require the presence of a viral trans-acting factor, which on its own may lead to transformation of the transfected cells. In particular, the large T antigen of the Polyomavirinae is known to have a broad interaction range, which has limited the exploration of BKV vectors as an efficient tool for gene therapy. At physiological levels, BKV T antigen is able to bind most of p53 in the cell . p53 is important in cell cycle arrest and apoptosis, which means that loss of p53 function is a major cause (50%) of transformation. Furthermore, it has been shown that BKV is mutagenic in human cells and that BKV T antigen induces chromosomal aberrations in human embryonic fibroblasts. Studies with SV40 large T antigen, which has high sequence and function homology with BKV T antigen, indicate that the large T antigen influences cellular gene expression by altering mRNA levels of cellular transcription factors, but also by interacting with and regulating the DNA-binding or transcriptional activity of specific transcription factors [180–186]. Therefore, further improvement of these vectors has mainly focused on the elimination of the T antigen. SV40 ori-based vectors, in which the large T antigen is exchanged for the matrix attachment region fragment of the human interferon β-gene, indicate that such a vector remains as an episome in the cell line CHO over more than 100 generations without selective pressure [187,188].
Also the transforming proteins E5, E6 and E7 of BPV-1 may interact with the cell’s own proteins and alter the normal cellular environment. It has, for example, been shown that the human analog of E7 interacts with p53. Furthermore, Uemura et al.  used BPV-1 gene products for activation of ras, and it has even been suggested to use this BPV-1-derived vector to test the transforming ability of oncogenes with relatively low oncogenic potential. Therefore, the BPV-1 genome was also altered to obtain a vector with decreased transforming capacity. Deletion of the transforming genes caused in some cell lines a decrease in plasmid copy number, while in most cell lines tested the episomal maintenance of this vector and its ability to express cDNA were retained. Furthermore, this vector could be stably maintained in several human cell types [58,81]. Although E5, E6 and E7 are the main transforming proteins, it has to been taken into account that some transforming features have also been attributed to E2 .
The oncogenicity of EBV is mainly ascribed to EBNA2  and LMP1 [189–190]. EBNA2 is the earliest EBV gene to be expressed after infection. The role of EBNA2 in latency and pathogenesis probably centers around its ability to activate both viral and cellular promoters . EBNA1 has long been reported to possess no transformation activity. As EBNA1 is the only viral protein produced in EBV vector-transduced cells, EBV-based vectors would not be oncogenic and were presented as a safe gene-delivery system. It was shown that, in a hepatic background, oriP plasmids were able to replicate stably and that neither high levels of EBNA1 nor multiple episomes interfere with the expression of liver-specific proteins . On the other hand, in vitro experiments have suggested that EBNA1 may not be as innocent as previously thought. It can, for example, bind to RNA, at least in vitro, and may therefore be capable of influencing expression at the post-transcriptional level . In addition, Sung et al.  showed some functional similarities between the EBV oriP and the enhancer region near the c-myc gene, raising the possibility that, under particular circumstances, this sequence could also interact with EBNA1, resulting in deregulation of the proto-oncogene c-myc. In vivo experiments also create doubts about the harmlessness of EBNA1. The expression of EBNA1 predisposes B cells to lymphoma in transgenic mice. The B cell tumors are remarkably similar to those induced by transgenic c-myc expression. These results demonstrate that EBNA1 has oncogenic features in mice . However, there are also reports demonstrating the safety of this vector in in vivo experiments. No pathological changes were found in mouse liver transfected with the EBV replicon vector by liposomes of hemagglutinating virus of Japan .
Episomal hybrid vectors
As each system has its strengths and weaknesses, hybrid vectors have been created that combine features of different viral systems. An interesting characteristic of SV40 vectors is that they become highly amplified in COS cells (e.g. 200 000 episomal copies per cell within 48 h of transfection) in short-term or transient expression experiments. In this case, a direct correlation is found between the level of T antigen synthesis and the extent of episomal replication of plasmids containing an SV40 origin . Although the SV40 vector system offers a rapid and efficient way to introduce and express foreign DNA into cells [194,195], it also has limitations. First, the host range is limited to certain primate cells. Secondly, transfection of most permissive cells with recombinant SV40 ori-containing vectors culminates in death, thus limiting studies to transient periods. One solution to this problem is to generate special cells in which the vector copy number can be regulated by the expression of the SV40 T antigen. Therefore, episomal expression plasmids were constructed that combine the advantages of episomal vectors, such as those derived from BPV-1 or EBV, with unique features of the SV40 origin, i.e. rapid amplification of the plasmid. To this end, an SV40 ori was inserted into episomal BPV-1-derived and EBV-derived vectors in which SV40 replication is dominant . On addition of SV40 T antigen, e.g. after transient transfection of the cells with an SV40 T antigen-expressing plasmid, the episomal vector replicates to a high copy number, reaching 100 000 copies per cell [196,197]. These vectors were extremely promising for transient protein overproduction. However, a high mutation frequency, especially recombination, is often associated with replication from the SV40 ori on the EBV–SV40-based shuttle vectors, whereas EBV replication itself induces no detectable increase in mutation frequency [121,198,199].
Most of the current gene therapy approaches make use of viral vectors because of their specific capacity to deliver DNA to cells. However, the lack of sustained expression as well as host immune reactions are serious drawbacks of these systems. Episomal vectors can provide such sustained expression without immune reactions, although efficient gene delivery remains a problem. Therefore, hybrid vectors were developed consisting of EBV sequences (EBNA1 and oriP) and retroviral , adenoviral  or herpes viral  elements. An EBV/retroviral hybrid vector expressing green fluorescence protein showed an infection efficiency of 50% of hemopoietic progenitor cells . However, no data are yet available to validate this vector system for in vivo gene therapy experimentation. The use of a combinatorial EBV/adenoviral hybrid vector, in which the E1 gene of adenovirus was deleted, also resulted in high transformation efficiency (37%) of D-17 cells . The vectors were episomally maintained for at least 14 weeks, although some chromosomal integration was noticed. At maximal transformation efficiency, strong toxicity to the cells was observed, which currently limits the wide application of this system. The use of ‘gutless’ adenoviral vectors may improve it. So far, the most promising results have been obtained with a combined EBV/herpes virus vector . This vector showed efficient transgene expression in a variety of human cell lines in vitro; it is, moreover, capable of carrying large inserts of DNA and persists for at least 2–3 months in vivo. The results are promising and demonstrate efficient delivery and long-term retention of the vector in vitro. However, more in vivo studies are necessary to prove the long-term applicability and safety of such vectors.