Promoter from Chinese hamster elongation factor-1a gene and Epstein-Barr virus terminal repeats concatemer fragment maintain stable high-level expression of recombinant proteins

Creation of p1.1-D1, p1.1-D2, and p1.1-D expression plasmids with different lengths of EEF1A1 DFR

p1.1-eGFP: The control plasmid p1.1, described in detail in Orlova et al. (2014), was deposited to Addgene as a plain vector (#162737) and with cloned fluorescent reporter proteins: p1.1-eGFP (#162738), p1.1-mCherry (#162772).

p1.1-D1-eGFP (Δ676 DFR) was obtained by restriction of the p1.1-eGFP with BstZ17I enzyme and self-ligation.

p1.1-D2 (Δ2,715 DFR) was obtained by BamHI restriction of the plasmid with the cloned EEF1A1 DFR pAL-3CH (Orlova et al., 2014) followed by self-ligation. After that, the truncated EEF1A1 DFR (Orlova et al., 2014) was cut at the XhoI and NheI sites and ligated with the pBL-ID-EBV-5CH plasmid (Orlova et al., 2014) cut at the SalI and XbaI sites. During this procedure, 2,715 bp were removed from the 4,264-bp DFR, and 1,549 bp of DFR remained. The resulting p1.1-D2 vector plasmid was inserted with the eGFP gene using AbsI and NheI, yielding p1.1-D2-eGFP.

p1.1-D (Δ3,644 DFR), the mutant with the shortest DFR, was created using plasmid pAL-3CHEF-F1 (Orlova et al., 2014), which was cut by NheI and XbaI. The DFR fragment was ligated into pBL-ID-EBV-5CH, restricted with XbaI enzyme, and the required orientation was selected using PCR. The resulting p1.1-D plasmid was inserted with the eGFP gene using AbsI and NheI, yielding p1.1-D-eGFP.

Creation of p1.1-Tr1, p1.1-Tr2, and p1.1-Tr3 expression plasmids with different lengths of EEF1A1 UFR

Three fragments of the EEF1A UFR region flanked by the EcoRI and FseI adaptor restriction sites were obtained using PCR and p1.1 as a template. The PCR-products were subcloned into the pAL2T vector and sequenced. The plasmid p1.1-D-eGFP, 8,995 bp in size, was cut by EcoRI and FseI enzymes, and the PCR-generated truncated variants were inserted in place of the 2,845-bp EcoRI-FseI UFR fragment.

p1.1-Tr1-eGFP (Δ3,644 DFR, Δ1,608 UFR): Using the primers AD-5CHD2083-EcoRIf and AD-5CHD-FseIr (Table 1), a 2,083-bp PCR product was obtained, cloned into pAL2T T-vector, sequenced with primers M13 and SQ-DEL2083-R (Table 1), and then a 2,076-bp EcoRI-FseI fragment was cut and transferred into the p1.1-D-eGFP plasmid restricted with the same enzymes. The total size of the preserved UFR (EcoRI-AbsI) was 3,309 bp.

p1.1-Tr2-eGFP (Δ3,644 DFR, Δ2,400 UFR): Using the primers AD-5CHD1291-EcoRIf and AD-5CHD-FseIr, a 1,291-bp PCR product was obtained, cloned into the pAL2T vector, sequenced with primers M13 and SQ-DEL1291-R (Table 1), and then a 1,284-bp EcoRI-FseI fragment was cut and transferred into the p1.1-D-eGFP plasmid restricted with the same enzymes. The total size of the preserved UFR (EcoRI-AbsI) was 2,517 bp. The resulting p1.1-Tr2-eGFP plasmid is available from Addgene, plasmid #162782.

p1.1-Tr3-eGFP (Δ3,644 DFR, Δ3,264 UFR): Using the primers AD-5CHD427-EcoRIf and AD-5CHD-FseIr, a 427-bp PCR product was obtained, cloned into the pAL2T vector, sequenced with primers M13 and SQ-DEL427-R (Table 1), and a 420-bp EcoRI-FseI fragment was cut and transferred into the p1.1-D-eGFP plasmid restricted with the same enzymes. The total size of the preserved UFR (EcoRI - AbsI) was 1,653 bp.

Creation of expression plasmids to evaluate the functional role of EBVTR

p1.1-Tr2(-)EBVTR-eGFP was obtained by the p1.1-Tr2-eGFP restriction with EcoRI and PvuI The 959 nt fragment containing EBVTR and a part of the ampicillin resistance gene was deleted and changed to the 551 nt EcoRI-PvuI fragment of the p1.1(EBVTR-) plasmid (Orlova et al., 2014) containing no EBVTR.

pOptiVEC-eGFP was obtained on the basis of the pOptivec-circ plasmid (self-ligated pOptivec-TOPO, Thermo Fisher Scientific, Waltham, MA, USA), which was restricted withBamHI-NotI and ligated to the BamHI-NotI fragment of the pEGFP-N2 (Clontech, Mountain View, CA, USA), which contained the eGFP ORF.

pOptiVEC(+)EBVTR-eGFP was obtained by restriction of the pOptivec-circ with SpeI and PvuI enzymes and ligation with the EBVTR-containing NheI-PvuI fragment of the pBL2-ID-EBV (Orlova et al., 2014). The eGFP gene was cloned by BamHI-NotI.

The specific primers used for cloning are listed in Table 1. We used restriction endonucleases from Sibenzyme (Novosibirsk, Russia) and oligonucleotides, PCR reagents, and nucleic acid purification kits from Evrogen (Moscow, Russia). Cloning was performed in the Escherichia coli TOP10 strain (Invitrogen, Carlsbad, CA, USA), and competent cells were prepared in-house. DNA sequencing was carried out at the Interinstitutional Center for Collective Use “GENOME” IMB RAS. The genetic maps of all the plasmids are provided in Data S1.

CHO DG44 cell culture

A DHFR^−/− CHO DG44 cell line (Thermo Fisher Scientific, Waltham, MA, USA) was used. The untransfected cells were cultured in 125 mL shake flasks with 30 mL of ProCHO5 medium (Lonza, Basel, Switzerland) supplemented with 4 mM glutamine, 4 mM alanyl-glutamine, and HT (in the form of 500 µM sodium hypoxanthine, 80 µM thymidine, all supplements from PanEco, Moscow, Russia) at 37 °C in a 5% CO₂ incubator. The cells were passaged every 3–4 days. Cell density and viability were determined using the trypan blue exclusion method and a haemocytometer.

Transfection of CHO DG44 cells

Electroporation was performed using the Neon Transfection System (Thermo Fisher Scientific). Fifty micrograms of each plasmid were ethanol-precipitated, dried under laminar air flow, and dissolved in 50 µL of R-buffer (Thermo Fisher Scientific). The cells were split 24 h before transfection. Ten million cells were washed once in Dulbecco’s phosphate-buffered saline (DPBS), resuspended in 50 µL of R-buffer, and gently mixed with the appropriate plasmid DNA, resulting in a total volume of 100 µL. The whole volume was placed in the electroporation tip and pulsed once (1,700 V, 20 ms). The transfected cells were transferred into a 125 mL Erlenmeyer flask containing 30 mL of prewarmed ProCHO5 medium supplemented with 8 mM glutamine and HT. The transfection efficiency was assessed 48 h after transfection as the GFP fluorescence in relation to viable cell density.

Selection of stable transfectants and genomic amplification of the transgene

Selection was performed by replacing the transfection medium with ProCHO5 medium containing 200 nM MTX. The cells were passaged every 3–5 days with centrifugation (300 g, 5 min) until the cell viability was restored to 80%. One-step genomic amplification was driven by increasing MTX concentration to 2,000 nM, and the cells were passaged until the cell viability was restored to 80% (typically approximately 15 days or five passages).

Samples of stably transfected and amplified cells were collected for eGFP expression level analysis (flow cytometry and measurement of eGFP concentration in cell lysates).

Stability studies

Stability studies were conducted for 2,000 nM MTX-amplified cultures, unless stated otherwise. To assess the long-term stability of the eGFP expression level, the cells were cultured for two months (63 or 68 days) in the absence of selection pressure with 3–4 day passages using centrifugation (300 g, 5 min), seeding density 3  × 10⁵ cells/mL. At every two or three passages, cell samples were collected for eGFP expression level analysis (flow cytometry and measurement of eGFP concentration in cell lysates).

Flow cytometry

The cells were analyzed during the exponential growth phase using a NovoCyte flow cytometer (Agilent, Santa Clara, CA, USA). Data were acquired for at least 50,000 individual cells using 488 nm excitation with a 530/30 nm bandpass filter to detect eGFP and analyzed using NovoExpress software (Agilent).

Measurement of eGFP concentration in cell lysates

The cell culture samples containing approximately 5 × 10⁶ live cells were centrifuged at 300 g, and the pellets were washed once with PBS and stored at −20 °C for further analysis. The pellet was resuspended in 100 µL of lysis buffer containing 100 mM Tris-HCl pH 8.0, 100 mM sodium chloride, 10 mM EDTA pH 8.0, 0.5% (v/v) Triton X-100, and 1x Protease Inhibitor Cocktail (Sigma, Burlington, MA, USA); vortexed; incubated for 5 min at RT; and then centrifuged for 2 min at 13,400 rpm. The supernatant was transferred into a new tube and centrifuged for 5 min at 13,400 rpm, and the lysate was transferred to a new tube and stored at −20 °C for further analysis.

The enhanced GFP content in the lysates was evaluated by measuring fluorescence intensity at an emission wavelength of 509 nm using the Fluorat-02-Panorama spectrofluorometer (Lumex, St.Petersburg, Russia). An external calibrator, rTagGFP2 (Evrogen), was used as a fluorescence standard. Total protein content in the lysates was measured using the Bradford method with Brilliant Blue Protein Reagent (Sigma). The eGFP expression level was calculated as the eGFP concentration divided by the total protein concentration in the cell lysates.

Quantitative real-time PCR

The transgene copy number in the genomic DNA was evaluated using qPCR. Genomic DNA was isolated from 2.5 × 10⁶ cells using the Wizard SV Genomic DNA Purification System (Promega, Madison, WI, USA) and quantified using a Qbit fluorimeter (Thermo Fisher Scientific), Qbit DNA HS kit, and 5 ng of DNA for each reaction.

For eGFP and DHFR genes, multiplex qPCR was performed using the qPCRmix-HS UDG reaction mixture (Evrogen) and two combinations of primers and fluorescent probes (Table 1), with the p1.1-eGFP plasmid as a standard. For the Rab1 gene, the qPCRmix-HS SYBR reaction mixture (Evrogen) and the pGEM-Rab1 plasmid were used.

Primers and probes used for qPCR were purchased from Lumiprobe (Moscow, Russia). The iCycler iQ thermocycler (Bio-Rad, Hercules, CA, USA) was used for amplification and detection. The cycling parameters were as follows: initial denaturation for 10 min at 95 °C and 40 cycles of amplification (10 s denaturation at 95 °C, 15 s annealing at 55 °C, 15 s elongation at 72 °C). Each sample was analyzed in triplicate. The iCycler Iq4 program was used to determine threshold cycles, PCR efficiency, and copy numbers. The weight of one CHO haploid genome was 3 pg, according to (Gregory et al., 2007).

Statistical analysis

T-tests were performed using Prism (GraphPad Software, San Diego, CA, USA), available at https://www.graphpad.com/quickcalcs/ttest1.cfm (accessed on March 18, 2023).

Effects of DFR deletions

According to data from Deer & Allison (2004), the complete removal of the 4.2 kbp-long DFR of the EEF1A1 gene from the expression vector resulted in a significant (1.7- to five-fold) decrease in the target protein expression level. The full-length vector p1.1 employed by our group differed from the vector pDEF38 used in the initial study in that it contained the EBVTR genetic element, which was not expected to affect the transcriptional stability of the target gene. The DFR of the Chinese hamster EEF1A1 gene in the plasmids pDEF38 and p1.1 contained two Alu-like repeats (Haynes et al., 1981) with the predicted microRNA copies. To check their effect on target gene expression stability, we deleted one (p1.1-D1-eGFP, Δ676 bp) or both (p1.1-D2-eGFP, Δ2,715 bp), as shown in Fig. 1. The resulting plasmids were transfected into CHO DG44 cells, and the selection was performed at 200 nM MTX and stably transfected cell populations were studied in the simple batch culture for 80 days without MTX. All tested plasmids gave rise to polyclonal cell populations with similar eGFP expression levels at 200 nM MTX and demonstrated a similar slight drop in eGFP expression after 80 days of culture, less than a 20% decrease from the initial values (Fig. S1A). The copy number of the genome-integrated target genes, determined by qPCR analysis, dropped four to 10-fold on day 80 (Fig. S1B). This drop was consistent with observations made by other groups. Under similar conditions, a three-fold drop in copy number was observed for HEK293 dhfr- cells and plasmids with the DHFR marker (Lee, Baek & Lee, 2021). No correlation between copy number and specific productivity was observed (Lee, Baek & Lee, 2021).

The deleted Alu-like repeats and supposed microRNA coding regions did not affect the target gene transcription level or stability of target gene expression. We observed a statistically significant 27% decrease in the amplification rate upon the deletion of the DFR. This effect is expected to be the consequence of the elimination of two Alu-like elements in the DFR that are known to contribute to the double-strand break reparation (Morales et al., 2015). The double-strand breaks occur in large numbers during MTX-driven target gene amplification (Baik, Han & Lee, 2021) and drive the amplification events. This effect of the amplification rate decrease was moderate and completely negated by the truncation of the UFR.

The deleted DFR region was considered expendable and completely removed from the plasmid vector. The resulting plasmid, p1.1-D-eGFP (Δ3,644 bp), was reduced to 9 kbp length and subjected to further deletions in the UFR.

Effects of UFR deletions

Three more plasmids (p1.1-Tr1-eGFP, p1.1-Tr2-eGFP, and p1.1-Tr3-eGFP) with deletions in the UFR were constructed and used together with two control plasmids (full-size p1.1-eGFP and p1.1-D-eGFP with the intact UFR and deleted DFR) (Fig. 1). CHO DG44 cells were transfected by electroporation, and the transfection efficiency was 20% for p1.1 and between 3% and 9% for other plasmids. Stably transfected cell populations were generated for all plasmids simultaneously in the presence of 200 nM MTX and subjected to one round of target gene amplification by increasing MTX concentration to 2,000 nM. MTX-driven genomic amplification is widely used for the generation of highly productive cell lines and is related to the increased overall genomic instability of cell lines and lower expression level stability of target genes due to unwanted chromosomal rearrangements through double-strand break repair (Baik, Han & Lee, 2021) and other effects. As shown in Fig. 2A, the eGFP expression level was nearly the same for all populations after the initial selection, but differences appeared after genomic amplification. The eGFP level was inversely correlated with the plasmid size. The eGFP expression level was the highest (4.57 ± 0.09% of total cellular protein) for the shortest plasmid (p1.1-Tr3) and was 3.25 ± 0.10% for the longest initial plasmid (p1.1). The amplification rate, calculated as the ratio of the eGFP content before and after the genomic amplification, was inversely correlated with the plasmid size for all truncated plasmids, but was higher than expected for the initial long p1.1 plasmid (Fig. 2B).

Gene-amplified cell populations were cultured in the absence of selection pressure and HT supplementation for 63 days. These culture conditions closely mimic the industrial process for producer cell lines, and the stability of the target gene expression rate for 60 cell generations is usually considered sufficient for master cell bank validation. We observed that in the absence of selection pressure, all cell populations gradually decreased eGFP expression (Fig. 2C), which coincided with the loss of a transgene inserted in the genome as measured by qPCR using primers for eGFP and DHFR ORFs (Fig. 2E). We performed long-term culture for the polyclonal populations, but not for clonal cell lines. Therefore, individual cells with a higher growth rate and lower content of recombinant protein will take over the population during the two months of culturing. Previously, we demonstrated that many clonal cell lines based on the p1.1 plasmid are stable for 60–90 days of cultivation in a non-selective medium (Orlova et al., 2017; Orlova et al., 2014; Orlova et al., 2019; Kovnir et al., 2018), so we considered the eGFP production drop for the p1.1-eGFP plasmid as a minimal possible decrease.

On day 63 of the long-term culture, there was a statistically significant productivity decrease for the p1.1-Tr3 plasmid in comparison to the p1.1-Tr2 plasmid, and the productivity decrease for the p1.1-Tr2 plasmid and all longer plasmid variants were indistinguishable from the full-size p1.1 control plasmid (Fig. 2C). The flow cytometry analysis data (Fig. 2D) were consistent with the fluorometric analysis of the cell lysates. Over time, minor subpopulations with a very low eGFP signal developed in all populations. Only the p1.1-Tr3 population also had a major subpopulation with a reduced eGFP signal, visible on the diagram as the left shoulder of the eGFP signal strength distribution peak. For the p1.1-Tr3, the highest percentage of cells (53.3%) fell into the area of fluorescence signal strength of the least productive 10% of cells (gate for the p1.1 population).

The production drop was similar for all truncated plasmids except for p1.1-Tr3. The shortest plasmid, p1.1-Tr2, had an acceptable target gene expression loss rate and was taken for further analysis.

Effect of the EBVTR fragment

The major difference between the p1.1 plasmid and the pDEF38 plasmid, developed by Deer & Allison (2004), is the presence of the 400-bp non-coding EBVTR element: the fragment of the EBV terminal repeats concatemer. EBVTR greatly increased the chromosomal integration rate of the p1.1-eGFP plasmid (Orlova et al., 2014). We hypothesized that the EBVTR element may also affect target gene expression. We removed it from the p1.1-Tr2-eGFP plasmid and compared the p1.1-Tr2-eGFP and p1.1-Tr2(-)EBV-eGFP plasmids by simultaneously preparing two stably transfected and gene-amplified cell populations, as described above. The level of eGFP expression was indistinguishable at both 200 nM and 2,000 nM MTX (Figs. 3A, 3B). The decrease in the eGFP expression level at 63 days of culture in a non-selective medium was higher for the (-) EBV plasmid, and the difference was statistically significant on day 54 (Fig. 3C). The flow cytometry data showed that at the end of the long-term culture, the pool containing the p1.1-Tr2 plasmid retained 69.6% of the initial mean fluorescence intensity versus 59.1% for the p1.1-Tr2(-)EBV plasmid (Fig. 3D). This change was visible in the histogram as the peak shift towards the lower fluorescence intensity. The percentage of the lowest expressing cells at the end of the stability run was roughly the same (41.1% versus 44.4%) for both plasmids. We observed the same transgene copy numbers for both plasmids at the end of the stability studies, but a higher eGFP level for the plasmid with EBVTR (Figs. 3C, 3E). Therefore, the EBVTR fragment was shown to affect expression level stability not by maintaining integrated transgene copies in the host genome, but by another mechanism.

To evaluate the effect of the EBVTR fragment, we inserted it into a commercially available pOptiVEC plasmid (Thermo Fisher Scientific) based on the CMV promoter. The resulting plasmids pOptiVEC(+)EBVTR-eGFP, pOptiVEC-eGFP, and the positive control p1.1-Tr2-eGFP plasmid (denoted below as pOV(+)EBV, pOV, and p1.1-Tr2) were simultaneously transfected into the CHO DG44 cells and tested as described above. We observed a statistically significant difference in eGFP expression level for the 2,000 nM MTX-amplified cultures: 3.62 ± 0.19% for p1.1-Tr2, 4.93 ± 0.18% for pOV(+)EBV, and 5.49 ± 0.20% for pOV. This demonstrates that the CMV-based plasmid gives a higher yield of the model protein after one-step transgene amplification (Fig. 4A). The amplification rate was also higher for smaller CMV-based plasmids: the increase in eGFP expression level was six-fold for p1.1-Tr2 (7.4 kbp), 14-fold for pOV (5.1 kbp), and 10-fold for pOV(+)EBV (5.5 kbp) (Fig. 4B). The CMV promoter-based plasmids are industry standard due to the great strength of the promoter, but they suffer from instability in their target gene expression levels. Unsurprisingly, in the case of the pOV plasmid, we observed a six-fold expression level drop after 68 days of culture in a non-selective medium, a two-fold drop in the case of the p1.1-Tr2 plasmid, and three-fold drop for the pOV(+)EBV plasmid (Fig. 4C). The flow cytometry data indicated that high-expressing cells completely disappeared in the case of the pOV plasmid but remained in the other two cases (Fig. 4D). Quantitatively, the area of 10% cells with the lowest eGFP fluorescence intensity (gate for the p1.1, day 0 in the long-term culture) on day 68 included 53.6% cells for the p1.1-Tr2, 73.9% of cells for the pOV(+)EBV, and 88.0% of cells for the pOV. According to the qPCR data (Fig. 4E), the pOV plasmid gave more genomic copies after genomic amplification than the other plasmids and most of these copies were retained during the culture in non-selective conditions. The resulting target gene expression level per gene copy was low, indicating promoter inactivation as the main reason for the productivity drop rather than transgene copy loss. The behavior of the pOV(+)EBV-based cell population was close to that of the p1.1-Tr2 population according to the flow cytometry data: highly-producing cells were mostly preserved, and the productivity drop coincided with the appearance of the low-producing subpopulation. Despite the superior stability of the pOV(+)EBV plasmid-based populations, this plasmid vector was not superior to the p1.1-Tr2 because the eGFP expression level after 68 days of culture in non-selective conditions was lower for the pOV(+)EBV than for the p1.1-Tr2.

Analysis of UFR in the orthologous human EEF1A1 gene

The UFR of the EEF1A1 gene deleted in the p1.1-Tr1, p1.1-Tr2, and p1.1-Tr3 plasmids were compared with a detailed map of the orthologous human EEF1A1 gene, compiled from the ENCODE project data (http://genome.ucsc.edu) (Kent et al., 2002). Figure 5 shows ChIP-seq-defined markers of chromatin transcriptional activity, including transcription factor binding sites (TFBS) and DNAase I hypersensitivity sites (DHS). The human EEF1A1 gene and its rodent orthologs have many homologous regions, but some regions upstream of the promoter are highly divergent. In these regions, the functional properties of hamster and human DNA can differ greatly.

The shortest version of the truncated p1.1 plasmids that we studied, p1.1-Tr3, closely matched commercially available expression vectors with human or hamster EF1 α core promoters. The truncated EEF1A1 UFR of the p1.1-Tr3 retained two complete DHS clusters out of five and one DHS fragment, two complete TFBS clusters and a cluster fragment, and one complete and one incomplete histone H3 Lys27 hyperacetylation site (H3K27ac site) out of five. Most of the functional sites identified by ENCODE for the EEF1A1 gene UFR were absent, which may explain the lack of expression stability for p1.1-Tr3 plasmid.

In the case of the p1.1-Tr2, three TFBS clusters, DHS, and the total number of H3K27a sites out of five located upstream of the EEF1A1 promoter remained intact. About 250 bp upstream of the EEF1A1 gene included in the p1.1-Tr2 were not homologous between humans and rodents. We assume that this truncated variant preserves the entire functionally significant UFR. Two more potential TFBS clusters located further upstream were removed from this plasmid variant.

The full-length EEF1A1 UFR, compared to the p1.1-Tr1, contained a possible mark of chromatin transcriptional activation - a DNA region with simultaneous increase in histone H3K27 acetylation and TFBS and DNAase I hypersensitivity. At the same time, significant homology for this DNA site was observed only between H. sapiens and M. musculus, but not between H. sapiens and R. norvegicus or between H. sapiens and C. griseus. Thus, it is likely that this region may have some equivalent in the Chinese hamster genome, but its exact position relative to the EEF1A1 promoter and its internal structure remains unknown.