The EF‐1α promoter maintains high‐level transgene expression from episomal vectors in transfected CHO‐K1 cells

Abstract In our previous study, we demonstrated that episomal vectors based on the characteristic sequence of matrix attachment regions (MARs) and containing the cytomegalovirus (CMV) promoter allow transgenes to be maintained episomally in Chinese hamster ovary (CHO) cells. However, the transgene expression was unstable and the number of copies was low. In this study, we focused on enhancers, various promoters and promoter variants that could improve the transgene expression stability, expression magnitude (level) and the copy number of a MAR‐based episomal vector in CHO‐K1 cells. In comparison with the CMV promoter, the eukaryotic translation elongation factor 1 α (EF‐1α, gene symbol EEF1A1) promoter increased the transfection efficiency, the transgene expression, the proportion of expression‐positive clones and the copy number of the episomal vector in long‐term culture. By contrast, no significant positive effects were observed with an enhancer, CMV promoter variants or CAG promoter in the episomal vector in long‐term culture. Moreover, the high‐expression clones harbouring the EF‐1α promoter tended to be more stable in long‐term culture, even in the absence of selection pressure. According to these findings, we concluded that the EF‐1α promoter is a potent regulatory sequence for episomal vectors because it maintains high transgene expression, transgene stability and copy number. These results provide valuable information on improvement of transgene stability and the copy number of episomal vectors.


Introduction
Expression vectors play important roles in gene therapy. Vectors currently used for gene expression have a number of limitations [1][2][3], such as random integration into the host genome or only transient retention. Random integration may lead to insertional mutagenesis and silencing of the transgene. Therefore, the ideal vector, especially for gene therapy, should be retained in cells without integration. Nonviral episomal plasmid vector pEPI was first developed by Piechaczek et al. [4]. pEPI requires a scaffold/matrix attachment region (S/MAR) linked to an expression unit [5,6]. Active transcription upstream of the S/MAR running into this sequence is required and is probably sufficient for episomal replication. When S/MAR is placed upstream of a transcription termination site or is deleted, the vector is lost or gets integrated [7]. The pEPI vector replicates independently at a low copy number in all verified cells. In addition to the low copy number, other drawbacks of the pEPI vector include unstable or low transgene expression [8,9]. In the subsequent studies, considerable efforts were made to improve pEPI-based vectors, such as insertion of regulatory cis-acting elements and reducing the amount of bacterial sequence [10,11].
In our previous study, we constructed an episomal vector harbouring a 387-bp DNA sequence comprising a characteristic MAR motif and evaluated different promoters in the episomal vectors. We found that the CMV promoter performed the best in terms of expression magnitude and stability [11,12]. However, expression of the [Correction added on 12th July 2017, after first online publication: the author name Dr. Zhongjie Xu has been amended in this version to reflect the correct spelling.] transgene was unstable, the copy number was low, and the expression magnitude was still unsatisfactory.
Composition of a plasmid vector, including promoters, enhancer, polyadenylation signals and other expression elements, influences transgene expression magnitude and stability. Although the CMV promoter provides high gene expression, there are reports that with the extended cultivation periods, productivity decreases [13,14]. CMV is prone to transcriptional silencing that is associated with DNA methylation [15,16]. By contrast, several strong promoters have been exploited with the aim of long-term, potent expression of a gene. Promoters of mammalian origin have been used most widely for this purpose, among which the human eukaryotic translation elongation factor 1 alpha (EF-1a, gene symbol EEF1A1) promoter is constitutively active in a broad range of cell types. The EF-1a promoter is often active in cells in which viral promoters fail to express the controlled genes and in cells in which the viral promoters are gradually silenced [17][18][19]. Some studies have indicated that promoters of endogenous mammalian genes like EEF1A1 may be more resistant to silencing than viral promoters are [20][21][22]. The EF-1a promoter used in conjunction with flanking regions of the CHO EF-1a gene is more active in CHO cells as compared with the use of CMV and SV40 promoters alone [23,24].
To further enhance the function of promoters, various enhancer elements have been added upstream of the promoters. In this study, we evaluated performance of various enhancer elements, the EF-1a promoter, CAG promoter (a fusion promoter comprising a CMV enhancer, the chicken b-actin promoter and a rabbit b-globin splice acceptor site) and CMV promoter mutants in the episomal vector and tested their transgene expression level and stability in CHO-K1 cells. Our findings should benefit researchers designing episomal vectors to achieve high expression and longterm stability.

Construction of the vectors
Based on previously described vector pEM (Fig. S1A), the CMV promoter mutants (inclusion of cytosines at positions 404 and 542 that were point-mutated to guanosines, Fig. S1B) and three different enhancer elements added upstream of the promoters (Fig. S1C) were synthesized chemically by Sangon Biotech Co., Ltd. (Shanghai, China). The promoter and enhancer sequences are listed in Figure S2. The EF-1a and CAG promoters (Fig. S1D) were generated by polymerase chain reaction (PCR). To achieve directional cloning, Ase I and Nhe I restriction sites were introduced at the 5 0 ends of the primers. The PCR cycling conditions were as follows: four cycles of 95°C for 3 min., 60-56°C for 30 sec., 72°C for 40 sec.; followed by 20 cycles at 94°C for 40 sec., 55°C, 72°C for 40 sec.; and a final extension step at 72°C for 3 min. The PCR products were recovered and their sequences were confirmed, followed by digestion with Ase I and Nhe I (Takara Biotechnology Co., Ltd., Dalian, China). The products were then ligated into the pEM vector to obtain vectors containing the EF-1a or CAG promoter ( Fig. S1D and E). The schematics of the constructs are shown in Figure S1.

Cell culture and transfection
Chinese hamster ovary K1 cells (CHO-K1) (provided by the Institute of Laboratory Animal Sciences, Beijing, China) were cultured in Dulbecco's modified Eagle's medium (Gibco, Carlsbad, CA, USA) supplemented with 10% of foetal bovine serum (Gibco, Grand Island, NY, USA) and 1% of a penicillin-streptomycin solution (Solarbio, Beijing, China) in a humidified incubator at 37°C and 5% CO 2 . The cells were cultured at a density of 8 9 10 4 /well in 12-well plates. After the cells reached 80-90% confluence, triplicate transfections were performed for each vector using Lipofectamine â 2000 (Invitrogen, Waltham, MA, USA) or Poly-Jet TM (SignaGen Laboratories, Rockville, MD, USA), according to the manufacturer's instructions. Stably transfected cells were selected 48 hrs after transfection by the addition of Geneticin (G418; Invitrogen) to the medium at a concentration of 500 lg/ml with selection for 2 weeks. The selected cells were subcultured into 96-well plates to obtain monoclonal clones using the limiting dilution method. Stable clones were then cultured in either the presence or absence of G418 (250 lg/ml).

Flow cytometry
CHO-K1 cells were obtained by screening at different time-points after the transfection. eGFP gene expression was examined under an Olympus IX71 fluorescence microscope (Olympus, Tokyo, Japan). Microscope settings were as follows: applicable mirror unite U-MN, excitation maximum 488, emission maximum 507 and exposure time 100 ms. The percentage of enhanced green fluorescent protein (eGFP)-expressing cells (48 hrs post-transfection) and eGFP mean fluorescence intensity (MFI) of each sample were analysed using a FACSCalibur cytometer (Becton Dickinson, Franklin Lakes, NJ, USA). FACS settings were as follows: FSC voltage 60, SCC voltage 300 and FITC voltage 380. The percentage of eGFP-expressing cells was determined by eGFP antibody labelling. Briefly, CHO-K1 cells were digested with trypsin and collected, followed by resuspending in 100 ll of a mouse anti-GFP antibody solution (ZSGB-Bio, Beijing, China), and finally, the cells were analysed on the flow cytometer. The numbers of eGFP-expressing and eGFP-non-expressing cells were calculated according to the flow cytometer results, and the transfection efficiency was calculated as the ratio of the number of eGFP-expressing cells to the total cell number.

Stability testing
Cells were passaged in 6-well plates and cultured further. The MFI for each vector type was measured using the FACSCalibur cytometer, and the retention of eGFP expression for each vector was calculated as the ratio of the MFI values at the end to those at the start of stability testing. The retention rate was calculated in accordance with the eGFP protein levels with or without G418 in CHO-K1 cells after 30 generations, as compared with the original eGFP expression levels. The retention rate is defined as MFI of the original monoclonal cells/MFI of monoclonal

Plasmid rescue experiments
A modified Hirt protocol [11,25] was used to isolate extrachromosomal DNA from CHO-K1 monoclonal cells transfected with the pEMEa vector. Escherichia coli (Sangon Co., Ltd., Shanghai, China) was electroporated with the Hirt extract from approximately 10 6 stably transfected cells. E. coli transformants were selected using agar plates containing 100 lg/ml kanamycin. Plasmid DNA was prepared from individual resistant clones, digested with Nhe I or Nhe I/Ase I and visualized on 1.0% agarose gels, with the aim of verifying the episomal status of pEMEa vectors in CHO-K1 cells.

Southern blotting
For this analysis, genomic DNA and extrachromosomal DNA plasmid (from a Hirt extract involving 10 7 CHO-K1 cells transfected with pEMEa) were isolated, digested with the single-cutting restriction enzyme Ase I, separated on 0.7% agarose gels (20 V, 20 mA overnight) and blotted onto Amersham Hybond-N+ paper, according to the manufacturer's instruction (GE Healthcare, Buckinghamshire, UK). Finally, an eGFP probe was labelled with Digoxin (Roche, Mannheim, Germany). The hybridization was performed in Church buffer (0.25 M sodium phosphate buffer pH 7.2, 1 mM EDTA, 1% of BSA and 7% of SDS) at 65°C for 16 hrs.

Fluorescence in situ hybridization (FISH)
Fluorescence in situ hybridization was performed to determine the episomal sites and gene copy number of the pEMEa vector. Monoclonal cells transfected with pEMEa and showing different expression levels were collected. eGFP served as a probe and was labelled using a Digoxigenin-Nick translation kit (Roche, Mannheim, Germany). Samples were counterstained with 1 lg/ml 4 0 ,6 0 -diamidino-2-phenylindole before examination under a Leica DMRB fluorescence microscope with a Leica DC 300 f camera. Approximately 50 visual fields were examined, and mean copy numbers were calculated. Fifty metaphase plates were analysed by FISH for each vector type.

Statistical analysis
All experimental data were analysed in the SPSS 18.0 software (SPSS Inc., Chicago, IL, USA). Data are reported as mean AE standard deviation (S.D.). A post-analysis of variance, multiple comparison procedure, was performed next to assess pairwise differences in expression confirmed by analysis of variance. Differences with P values <0.05 were considered statistically significant.

Stability of expression of the transgene
At 48 hrs after transfection, stably transfected cells were selected by the addition of Geneticin (G418) to the medium at a concentration of 500 lg/ml with selection for 2 weeks. During this selection process, the differences between various vectors were reflected in the number of clones and eGFP expression magnitude (Fig. 2). The EF-1a promoter yielded the most number of clones and strongest eGFP expression as compared with all the other regulatory elements ( Fig. 2A, D, G). After 2 weeks, eGFP expression in CHO-K1 cells transfected with the plasmid carrying the EF-1a promoter was significantly higher than that in the cells transfected with the plasmid containing the CMV promoter (Fig. 2B, E, H), and little eGFP expression was observed in cells harbouring each of the other six vectors (a representative example is the plasmid containing the CAG promoter, Fig. 2C, F, I).

Evaluation of the EF-1a promoter in stably transfected CHO-K1 cells
Given that the EF-1a promoter performed best in terms of expression levels and clone numbers, we chose CHO-K1 cells transfected with the EF-1a promoter-carrying vector (pEMEa) for our subsequent experiments. The cells were subcultured into 96-well plates to obtain monoclonal cultures, and FACS analysis of CHO-K1 clones carrying pEMEa revealed uniform expression levels (Fig. 3). We selected three groups of monoclonal cells: group A showed high expression (the eGFP expression level was 1.35 9 10 5 to 2.30 9 10 5 MFI, e.g. clone #12, Fig. 3B); group B showed medium expression (the eGFP expression level was 3.53 9 10 4 to 4.86 9 10 4 , e.g. clone #10, Fig. 3C); and group C had low expression (the eGFP expression level was 4865-10002, e.g. clone #19, Fig. 3D). Compared with the negative control, the average difference in eGFP expression for high-, mediumand low-expression clones was 358.3-, 82.6-and 16.5-fold, respectively. The average eGFP level in the high-expression clone was 4.8, 21.1-fold higher than that in the medium-expression and the lowexpression clones, respectively (Fig. 3E).

Plasmid rescue experiments
Using the Hirt DNA extraction method, we extracted extrachromosomal DNA from cells of clone #12, 10 and 19 carrying pEMEa, and the isolated plasmids were identified by digestion with the restriction enzymes. DNA was isolated from 10 clone using the Hirt extraction method, transformed into E. coli and subjected to digestion with Nhe I or Nhe I /Ase I. For all three clones, the length of the plasmid DNA from CHO-K1 cells was found to be identical to the original vector DNA (one example is shown in Fig. 4A). Ase I /Nhe I digestion was expected to yield 1335-bp and 4500-bp fragments, whereas digestion with Nhe I was expected to produce a 5835-bp DNA fragment, indicating that the pEMEa vector was not integrated into the host cell's genomic DNA and existed episomally in CHO-K1 cells.

Southern analysis
Total DNA isolated from the Hirt extract was digested with Nhe I and then subjected to Southern blotting. From pEMEa-carrying clone #12, 10 and 19, a 5835-bp restriction fragment identical to the original pEMEa plasmid DNA was obtained (Fig. 4B). The Southern analysis and plasmid rescue experiments further supported the episomal status of plasmid pEMEa in CHO-K1 cells.

FISH analysis
We next evaluated episomal vector replication and analysed copy numbers by FISH analysis, which was performed on spread chromosomes of CHO-K1 cells transfected with pEMEa (clones #12, 10 and 19). This analysis revealed that the observed mitotic stability of the vector was a result of its episomal presence on metaphase spreads (Fig. 5A-D). Fifty metaphase spreads were analysed by FISH for each clone. An average vector copy number of 7.56 AE 3.18 was estimated in CHO-K1 cells in the high-expression group (clone #12; range, 1-12 copies per cell; Fig. 5B, E); in the medium-expression group, the copy number was 4.37 AE 2.96 (clone #10; range, 1-11 copies per cell; Fig.  5C, E); and in the low-expression group, the copy number was 2.48 AE 1.03 (clone #19; range, 1-9 copies per cell; Fig.  5D, E).

Quantitative PCR
To study the relation between the expression levels and copy number of the episomal vector, the number of plasmid copies per cell for each clone was analysed by quantitative PCR. The high-expression clone was estimated to have a slightly greater copy number (6.23 AE 1.90), the medium-expression clone was found to contain 3.56 AE 1.18 copies/cell, while the low-expression clone showed a slightly lower copy number (1.75 AE 0.86; Fig. 6A). The results suggested that the expression level of eGFP was related to the gene copy number (Fig. 6B).

Long-term stability of expression of the recombinant protein
All stable monoclonal cells were cultured further in either the presence or absence of G418, and MFI was measured in the cells to assess stability of expression of the recombinant protein at passages 9, 13, 17, 21, 25 and 30 post-transfection. In single-cell clones, the eGFP levels decreased gradually over time. The eGFP expression in the high-expression group was higher than that in the mediumexpression and low-expression groups at passage 30 post-transfection ( Fig. 7A-C). The most stable expression was achieved in the high-expression group, which retained 86.45% of the original expression level by passage 30 post-transfection in the presence of G418 selection pressure (Fig. 7D), and retained 72.18% of the original expression level at passage 30 post-transfection in the absence of G418 selection pressure (Fig. 7D). In contrast, the low-expression group showed a decrease during the long-term culture: only 32.37% was retained at passage 30 post-transfection in the presence of G418 selection pressure and 12.69% in the absence of G418 selection pressure (Fig. 7D).
In agreement with the MFI results, eGFP gene expression was also evident in CHO-K1 by fluorescence microscopy after 30 generations in culture, both under selection pressure and in the absence of selection pressure (Fig. 8).

Analysis of the effects of transcription factor regulatory elements (TFREs) on transgene expression
Promoter activity is related to transcription factor-binding sites and TFREs. The distributions of seven TFREs (SP1, NF-jB, STAT, HSF, GATA, TEF and CEBP) were assessed for the EF-1a, CAG and CMV  promoters. The CEBP, NFjB, STAT, GATA and HSF TFREs were abundant in the EF-1a promoter ( Table 1). The CAG promoter did not contain CEBP and GATA TFREs, and the CMV promoter did not contain GATA TFREs. The findings led us to conclude that TFREs CEBP, GATA and HSF contribute to promoter activity the most.

Discussion
Cell lines combining high production and stability are important for recombinant protein production, particularly in relation to proteins for gene therapy. Successful generation of overproducing cell lines requires creation of cell clones expressing the recombinant protein at high and stable magnitude. In this study, we attempted to improve the pEM vector function in terms of transgene expression magnitude, copy number and long-term stability by inserting three different enhancer elements, mutating the CMV promoter, and using strong promoters CAG and EF-1a. We found that the use of enhancers, mutating the CMV promoter, and the use of strong promoters increased the transfection efficiency and transient expression of the recombinant protein. In particular, we observed increased long-term stability and copy number under the influence of the EF-1a promoter especially in the high-expression clones .
CMV is prone to transcriptional silencing, which is associated with DNA methylation [27,28]. Mammalian DNA is predominantly methylated at cytosine bases that are part of CpG dinucleotides [29,30]. In the CMV promoter, the cytosines at positions 404 and 542 were found to be methylated frequently [16]. To test whether removal of CpG sites can stabilize CMV promoter-driven gene expression, we point-mutated C 404 and C 542 to G and studied the effect of these mutations on long-term expression stability in transfected CHO-K1 cells. The results indicated that CMV promoter mutation increased only the recombinant protein expression transiently and did not affect long-term stability, which is inconsistent with the results of Benjamin et al. [16], who reported that a single mutation of C -179 to G can stabilize the production of a recombinant protein significantly under the control of the CMV promoter in stably transfected CHO cells. This discrepancy may be explained by the fact that Benjamin et al. used integrating vectors to construct the CMV promoter mutants, whereas we used an episomal vector. The episomal vector was constructed using a MAR element, and the transgenic expression level was related not only to the promoter but also to other regulatory elements of the vector.
An enhancer is a DNA sequence that can determine the temporal and spatial specificity of expression and increase a promoter's activity. According to one study [31], we synthesized three different enhancer elements and added them (separately) upstream of the promoter. The three different enhancer elements included combinations of NF-jB, E-box, GC-box and C/EBPa elements. The results showed that Enhancer-1 increased only the recombinant protein expression transiently, without an effect on long-term stability. Enhancers 2 and 3 did not yield any transcriptional enhancement. An enhancer is a cis-  acting element that can increase transcription activity only when combined with tissue-specific transcription factors and is related to its controlled promoter. Hence, we cannot rule out that the transcriptionpromoting effect of an enhancer relies, to some extent, on elements associated with the promoter in question. Transcription factors are different between transient expression and stable expression; and this state of affairs might have resulted in Enhancer-1 0 s functioning only in transient expression. In addition, the CMV promoter (589 bp long) contains its own enhancer elements, and a new enhancer upstream of the promoter may not necessarily increase the promoter activity any further.
Transgene expression stability and magnitude are influenced by various components of a plasmid vector, including the promoter. The EF-1a promoter is known as one of the strongest promoters in various mammalian cell lines [32], and the CAG promoter has been used frequently to drive strong gene expression in mammalian cells. However, in the present study, the CAG promoter could not maintain transgene expression stably during long-term culture as compared with the CMV promoter. Moreover, the EF-1a promoter showed the highest promoter activity and tended to be more stable in long-term culture, even in the absence of selection pressure on the transfected CHO-K1 cells. The activity of a promoter thus depends on many factors, such as genomic cis-acting sequences, cell line, type of vector and transcription factor-binding sites. Yang et al. demonstrated that the CAG promoter can drive transgene expression in chick embryo cells [33], whereas we transfected CHO cells. This difference may explain the weaker expression of the transgene under the control of the CAG promoter in transfected CHO cells. As mentioned above, promoter activity is affected by transcription factor-binding sites or TFREs. The TFREs of the promoters used in this study were analysed, and the results revealed that the CEBP and GATA Table 1 Locations of various transcription factor-binding motifs within the three promoters TFREs are abundant in the EF-1a promoter, but GATA is absent in the CMV and CAG promoters. These findings lead us to hypothesize that in episomal vectors, the CEBP and GATA TFREs contribute to the promoter activity the most.
Our results revealed that the EF-1a promoter is a potent regulatory sequence for episomal vectors and maintains high transgene expression. Plasmids containing the EF-1a promoter were found to replicate efficiently, stably and extrachromosomally in CHO-K1 cells. In addition, we found that the transgene expression magnitude of cells transfected with the EF-1a promoter-carrying vector is associated with the copy number: an elevated copy number caused higher expression.
In this study, we investigated the activity of various promoters and promoter-variant vectors in transfected CHO-K1 cells. However, only the CHO-K1 cell line was tested here, and the vectors should be evaluated in other cell lines and in vivo. In addition, other cis-acting elements in expression vectors need to be optimized further to develop a high-efficiency expression system for gene therapy.