Generation of homogeneous cell populations with tunable levels of transgene expression
Dominique Schlicht a, Carole Estoppey a, Julie Macoin a, Blandine Pouleau a,
Martin Bertschinger b,*
a Ichnos Biosciences, Epalinges, Switzerland
b Ichnos Sciences, La Chaux-de-Fonds, Switzerland

A R T I C L E I N F O

Keywords:
Stable cell line development Homogeneous cell populations Low expressing cell populations Tunable expression

A B S T R A C T

We describe here a vector construct to establish homogeneous cell populations expressing a recombinant gene of interest (GOI) at tuneable levels, including low expression levels that are difficult to generate using standard cell line development techniques. This is achieved using a tricistronic mRNA that contains an open reading frame for the gene of interest, a first internal ribosome entry site (IRES), an open reading frame for a fluorescent reporter protein (such as green fluorescent protein, GFP), a second IRES and an open reading for an antibiotic resistance gene (such as puromycin N-acetyl-transferase, PAC). The resistance gene allows convenient selection of stable cell populations. The fluorescent reporter protein allows convenient homogeneity and expression stability as- sessments of the cell line. The expression level of the GOI can be adjusted by using different start codons for the open reading frame. These alternate start codons will initiate the translation of the GOI with different efficiency, leading to cell populations expressing different levels of the GOI, and similar levels of the fluorescent reporter through the first IRES and the puromycin resistance gene through the second IRES to the GOI. Such cell pop- ulations are useful tools, for instance to assess the safety of potent targeted therapeutics, as they allow the simplified generations of homogenous cell populations with different levels of target protein expression between populations.

1. Introduction
Stable cell lines expressing transgenes are commonly used to produce biologics and as tools for in vitro and in vivo experiments. The generation of high expressing cell lines and the use of fluorescent proteins as re- porter proteins for cell line development has been reported already (DeMaria et al., 2007; Freimark et al., 2010; Mancia et al., 2004; Shi et al., 2014; van Blokland et al., 2007).
However, the expression level of recombinant proteins does not al- ways need to be maximized. Cell lines with less than maximal expression levels of a gene of interest (GOI) are useful to address biological ques- tions that require GOI expression levels like the natural levels of expression. Controlled or reduced expression may also be required for toxic proteins (Kaufman, 2000), that may otherwise harm the produc- tion cell or affect cell metabolism. Cell lines with different levels of surface expression of a GOI, for example, may be interesting tools to establish the threshold expression level of a membrane displayed target

protein for efficacy of an antibody in vitro or in vivo and are thus relevant for preclinical studies, such as toxicity evaluations, for highly potent bispecific antibodies like blinatumomab (Zhu et al., 2016).
There has been comparatively little work describing ways to develop cell lines with the homogeneous expression of low levels of a GOI. Usually random integration events in the host cell genome lead to cell populations with different expression levels due to the genetic envi- ronment of the integrated DNA. In addition to the high expressing cell populations that are usually selected, every cell line development campaign therefore generates cell populations with medium or low levels of expression.
Different published approaches to generate medium and low expressing cell populations include the use of weak promoters (Qin et al., 2010), impairment of expression through codon usage (Quax et al., 2015), the use of different starting codons (van Blokland et al., 2007), alternate splicing (Fallot et al., 2009; Lucas et al., 1996) or using inducible systems (Kaufman, 2000).

* Corresponding author.
E-mail address: [email protected] (M. Bertschinger).

https://doi.org/10.1016/j.jbiotec.2020.10.008

Received 27 April 2020; Received in revised form 24 September 2020; Accepted 6 October 2020
Available online 9 October 2020
0168-1656/© 2020 Elsevier B.V. All rights reserved.

D. Schlicht et al.

One major drawback of the methods described above, is that the selection of low expressing cell populations is far more challenging than the selection of medium and high expressing populations. The analytics normally used e.g. flow cytometry, ELISA may be closer to their limit of detection or quantification. The lower the required expression level and the lower the sensitivity of the method, the more this analytical limi- tation will affect homogeneity assessments, as remaining secondary populations or the instability of the selected population may be difficult to detect.
We present an expression construct that allows the generation of cell lines with homogeneous high, middle or low levels of transgene expression. The construct allowed us to isolate cell lines by screening for maximal expression of the fluorescent reporter protein, while the gene of interest (linked with the reporter by an internal ribosome entry site; IRES) is expressed with reduced efficiency by using non-ATG start co- dons with reduced translation initiation efficiency.
2. Materials and methods
2.1. Cloning of expression constructs
The Ichnos expression vector pT1 contains an expression cassette with a CMV (cytomegalovirus) promoter followed by an MCS (Multiple Cloning Site), the mRNA fragment coding for GFP (green fluorescent protein) and pac (including the IRES elements) and a SV40 poly(A) sequence. The sequence information of the expression cassette has been submitted as Supplementary File S1.
The gene coding for human CD38 (cluster of differentiation 38; UniProt accession number P28907) used in this experiment was ordered from Imagenes (Source Bioscience, Nottingham, UK). The coding sequence was first amplified on cDNA by PCR (polymerase chain reac- tion) using the primers GlnPr2046 and GlnPr2047 (all primers used in this work are listed in Table 1). A second PCR was done with the amplicon obtained in this PCR using the primers GlnPr1544 and GlnPr2047, adding an NheI restriction site to the 5’end and a NotI re- striction site to the 3’ end. This PCR product was gel-purified and cloned using NheI/NotI restriction sites into pT1 opened with the same enzymes and treated with CIP (calf intestinal phosphatase). This plasmid was referred to as pT1-hsCD38 and contained the start codon ATG and a consensus Kozak sequence.
The CD38 coding sequence was amplified from pT1-hsCD38 using the primers GlnPr2255 and GlnPr2258 introducing a non-consensus Kozak sequence and an NheI restriction site at the 5’end and a XhoI restriction site at the 3’ end. This PCR product was gel-purified and cloned using NheI/XhoI restriction sites into the pT1 plasmid opened with the same enzymes and treated with CIP.
The resulting plasmid was named pT1-hsCD38-ATG.The coding sequence was amplified on pT1-hsCD38 plasmid by PCR using the primers GlnPr2256 and GlnPr2258 for the sequence “ACCGTGG” and GlnPr2257 and GlnPr2258 for the sequence “ACCTTGG”. The resulting PCR products contained a NheI restriction site at the 5’end and an XhoI restriction site at the 3’ end. These PCR products were gel-purified and

Table 1

Journal of Biotechnology xxx (xxxx) xxx

cloned using NheI/XhoI restriction sites into the pT1 plasmid opened with the same enzymes and treated with CIP.
They resulting plasmid were named pT1-hsCD38-GTG and pT1- hsCD38-TTG.
The coding sequence of human PSMA (prostate-specific membrane antigen) was synthesized using amino acid information from the UniProt knowledgebase entry Q04609 (FOLH1_HUMAN, Isoform 1). Synthesized DNA was ordered from Eurofins with flanking NheI / XhoI restriction sites. After digestion with NheI / XhoI restriction enzymes, the digestion product was gel-purified and cloned in plasmid pT1, opened using the same enzymes and treated with CIP in order to yield plasmid pT1- hsPSMA-ATG.
PCRs using the forward primer GlnPr2670 (for the start codon GTG) or the forward primer GlnPr2669 (for the start codon TTG), respectively, and the reverse primer GlnPr2671 were used to amplify hsPSMA while changing the ATG start codon to GTG or TTG. The resulting PCR prod- ucts contained a NheI restriction site at the 5’ end and an XhoI restric- tion site at the 3’ end. These PCR products were gel-purified and cloned using NheI/XhoI restriction sites in the plasmid pT1 opened with the same enzymes and treated with CIP. They resulting plasmids were named pT1-hsPSMA-GTG and pT1-hsPSMA-TTG.
Maxipreps using standard commercial kits of the plasmids were used for transfection.

2.2. Transfections and cell line development
Suspension CHO-S (Invitrogen, Carlsbad, CA) cells were transfected with the DNA constructs coding for PSMA. For this purpose, exponen- tially growing CHO-S cells were seeded at a density of 2.0 106 cells
/mL in 10 mL of OptiMEM medium (Invitrogen, Carlsbad, CA) in tubespin bioreactor 50 vessels (“tubespin”, TPP, Trasadingen, Switzerland). A polyethylenimine (PEI)/DNA complex was generated by preparing 75 μg of PEI (polyplus-transfection, Illkirch, France) in 500 μl of 150 mM NaCl and 25 μg of DNA in 500 μl of 150 mM NaCl. Both solutions were mixed, incubated for 10 min and then added to the cells. After 5 h incubation at 37 ◦C under shaking (200 rpm), 10 mL of fresh culture medium (PowerCHO2 CD (Lonza, Basel, Switzerland) supple- mented with 4 mM glutamine) were added to the cell suspension and the cell suspension was distributed in two tubespins (10 mL each). Then the cells were incubated on a shaking platform at 37 ◦C, 5 % CO2 and 80 % humidity for 24 h. Suspension A20 cells (TIB-208, ATCC, Manassas, VA) were transfected with DNA coding for CD38 by electroporation using the NEON electroporation device (Thermo Fisher Scientific, Waltham, MA).
For this purpose, 200 μg of DNA were added to 40 106 exponentially
growing A20 cells, resuspended in 2 mL of electroporation buffer (Neon Transfection System Kit, Thermo Fisher Scientific). The cells were electroporated using optimized settings (3 pulses with a length of 10 ms
at 1450 V) and diluted to a final volume of 60 mL and a cell concen- tration of 0.66 106 cells/mL. The cell suspension was seeded in 3 T150 flasks (20 mL per T-flask) and incubated in a static incubator at 37 ◦C, 5
% CO2 and 80 % humidity for 24 h.
Following the incubation for 24 h, the transfected cells underwent a

Summary of primers (Operon (Ebersberg, Germany) or from Microsynth (Balgach, Switzerland)) used in PCRs.
Primer name Primer sequence

GlnPr2046 CTCTGGGTTCCAGGATCCACTGGTGCCAACTGCGAGTTCAGCCCG
GlnPr2047 CGACTGCGGCCGCTCATCAGATCTCAGATGTGCAAGATG
GlnPr1544 GATCGCTAGCCACCATGGAGACAGACACACTCCTGCTATGGGTACTGCTGCTCTGGGTTCCAGGATCCAC
GlnPr2255 TAAGATCGCTAGCCTTTATGTCGACAGACACACTCCTGCTATGGGTAC
GlnPr2258 CCGCGATCCTCGAGTCATCAGATCTCAGATGTGCAAGATG
GlnPr2256 TAAGATCGCTAGCCACCGTGGAGACAGACACACTCCTGC
GlnPr2257 TAAGATCGCTAGCCACCTTGGAGACAGACACACTCCTGC
GlnPr2669 GCCAAGCTAGCCACCTTGTGGAACCTCCTGCAC
GlnPr2670 GCCAAGCTAGCCACCGTGTGGAACCTCCTGCAC
GlnPr2671 GCCACCCTCGAGTCAGGCTACTTCGG

D. Schlicht et al.

single round of limiting dilution under selective conditions in order to generate pool populations of stable cell lines in 96 well plates. For this purpose, the transfected cells were thoroughly counted and diluted to 200 000 cells/ ml in fresh growth medium containing the selection antibiotic puromycin (Sigma-Aldrich, St. Louis, MO) at a concentration known to allow selective growth of stable cell lines only. The cell sus- pension was distributed in 96 well plates (100 μL/well). The 96 well
plates were transferred to a static incubator at 37 ◦C, 5 % CO2 and 80 %
humidity. One week after the limiting dilution, a volume of fresh se- lection medium (with puromycin) was added to each well and the plates were incubated at 37 ◦C in the static incubator for an additional week. After two weeks, growth of cells was observed in the wells of the 96 well plates and the cell populations were screened using a microplate reader (Synergie 2, BioTek, Winooski, VT) for GFP expression. The highest expressing cell populations were selected and further expanded to tubespins. As the flow cytometric analysis allowed the identification of

Journal of Biotechnology xxx (xxxx) xxx

target and effector cells without antibody treatment. Cytotoxicity was determined using a lactate dehydrogenase (LDH) kit (Promega). Briefly, the plates to be assayed were centrifuged for 5 min at 300 rpm. Super- natants were collected (50 μL) and transferred into 96-well ELISA plates. For the maximum release condition, target cells were submitted to 3 freeze/thaw cycles in 1.5 mL eppendorf tubes. Eppendorfs containing the lysed target cells were centrifuged for 5 min at 300 g and 3 50 μL of supernatant were harvested and transferred to 3 empty wells for reading as the maximum release. Reconstituted reagent was added (50 μL) to the culture supernatants and reactions were incubated for 30 min at RT in the dark. Stop buffer was added and the plates were read for optical density (OD) at 490 nm on a Synergy HT2-Spectrophotometer. The percentage of killing activity was determined using the following equation (equation 1):
SpR — SRTarget Killing (%) = x 100

homogeneous populations at this level, no additional efforts to ensure clonal derivation were found necessary.
2.3. Flow cytometric analysis
The flow cytometric analysis was performed using the Guava device (Merck Millipore) or the FACS Calibur (Becton Dickinson) with a detection antibody directly coupled with APC (anti-human CD38, eBioscience, at 1:100 dilution, Anti-human PSMA (FOLH1) APC, Bio- Legend, at 1:50 dilution). Cells were incubated for 20 min and washed once before flow cytometric analysis.
2.4. sABC assay (specific antibody binding capacity)
Non-transfected A20 cells and A20-CD38 cell populations were harvested and seeded at 100’000 cells/well in a U-bottom 96 well plate. 4 wells were prepared for each cell population (secondary antibody only, mouse Isotype control, anti-CD38 antibody and cells alone). After centrifugation, 100 μL of primary antibody (mouse anti-hsCD38 mono- clonal antibody (HIT2), Invitrogen) were added to the cells at a con- centration of 2 μg/mL and incubated for 20 min at 4 ◦C. The cells were then washed with FACS buffer (PBS/2.5 % FBS/10 % Versene). Two new wells were filled with 50 μL of beads (Dako, QIFIKIT). The plate con- taining the cells and the beads was centrifuged and 100 μL of the sec- ondary antibody (F(ab’)-Goat anti-Mouse IgG (H L), APC, Invitrogen) were added at a dilution of 1/50. The plate was incubated for 20 min at 4 ◦C in the dark. Beads and cells were then washed twice with 100 μL FACS buffer, before being resuspended in 100 μL 7-AAD (diluted 1:200 in FACS buffer). The acquisition was done using the Cytoflex (Beckman Coulter). The cytometric analysis was performed with the software FlowJo (BD). The APC geomean of the live cells were exported to calculate the numbers of receptors on cell surface according to the manufacturer instructions.
2.5. ReDirected lysis assay (RDL)
A hundred microliters of CHO-PSMA cells were seeded at 100’000 cells/well in a U-bottom 96 well plate. Cells were incubated with 50 μL of CD3ε/PSMA bispecific BEAT (Bispecific Engagement by Antibodies based on the T cell receptor) antibody (10, 2, 0.04 μg/mL) and control. The CD3ε/PSMA bispecific BEAT engages human CD3ε with the Fab arm (composed by Hc and Lc) and human PSMA with the scFv arm (composed by an Fc-scFv). Fifty microliters of effector cells (isolated human T cells from human blood from three different donors) were then added at an Effector-to-Target (E:T) ratio of 5:1. Each well was topped up to a final volume of 200 μl with assay medium (RPMI supplemented with 10 % FBS, 1 % glutamine, 1 % Pen-Strep, all items sourced from Gibco) and plates were incubated at 37 ◦C and 5 % CO2 for 48 h. Spontaneous killing of the target cells by the effector cells was evaluated using the wells containing only target cells and the ones containing the

SpR (sample release) corresponds to the OD of samples containing effector cells (isolated human T cells), target cells and treated with the BEAT. SR (spontaneous release) corresponds to OD of spontaneous release obtained in wells with target cells only. MR corresponds to OD of maximum release on target cells submitted to 3 freeze/thaw cycles.
3. Results
3.1. Generation of the constructs
Two different transmembrane proteins were used in this study, CD38 and PSMA. Tri-cistronic expression constructs were generated using two IRES (derived from Encephalomyocarditis virus), between the three open reading frames (Fig. 1). This allowed the expression of the GOI in the first position, the fluorescent reporter protein GFP in the second position and of the gene encoding the resistance against puromycin (pac) in third position on the same mRNA. In order to obtain different expression levels of the gene of interest (open reading frame 1), the start codon was modified to start with ATG, GTG or TTG. The cloning steps leading to the different constructs summarized in Table 2 are described in detail in the Material and Methods section.

3.2. Generation of stable cell lines
Three different stable cell populations expressing human CD38 were established for each of the constructs described in Table 2 using the mouse lymphoma cell line A20. The cells were amplified, transfected with the CD38 expression constructs listed in Table 2 and stable cell lines generated as described in Material and Methods. After two weeks of selection, 24 cell populations with the highest signal for GFP expression were identified by screening the 96 well plates using a microplate reader (data not shown). After further expansion, a second screening was per- formed using flow cytometry. Three cell populations with a histogram showing a narrow and homogenous GFP signal were selected.
The three cell populations expressing human CD38 transfected with the construct containing the ATG start codon showed a high expression of this protein on the cell surface. ATG is the start codon well recognized by the translation machinery (although the constructs used a non- consensus Kozak sequence). The three pool populations transfected with the construct using the GTG start codon showed a medium level of expression of this protein on the cell surface as this start codon is less efficient to initiate translation and so fewer proteins were expressed compared to the standard start codon. The three cell populations generated with the construct containing the TTG start codon showed the weakest staining for CD38. The TTG codon was found to be highly inefficient to initiate translation and only minor amounts of CD38 were produced (see Fig. 2, left panel).
These results were confirmed with a second gene of interest (human
PSMA) in Chinese hamster cells. Using the PSMA expression constructs

Table 2

Fig. 1. Schematic drawing of the tricistronic construct used in this study.

respectively (Fig. 3A). The differential expression was maintained for 30

Summary of all constructs generated.

days of continuous passaging in presence of the selective agent (Fig. 3B).

Plasmid batch

Protein

Construct details Sequence

Used

3.3. Redirected Lysis assays with cells expressing different levels of

number

expressed

Kozak Start codon

(start codon underlined)

in Cell line

transgene

pT1_hsCD38- ATG

hsCD38 Non- consensus

ATG TTTATGT A20

The cell populations expressing different levels of PSMA (see Fig. 2) were used in a redirected lysis assay as target cells and human T lym-

pT1-hsCD38-
GTG
pT1-hsCD38- TTG
pT1-hsPSMA- ATG
pT1_hPSMA- GTG
pT1_hsPSMA- TTG

hsCD38 Consensus GTG ACCGTGG A20

hsCD38 Consensus TTG ACCTTGG A20 hsPSMA Consensus ATG ACCATGG CHO hsPSMA Consensus GTG ACCGTGG CHO hsPSMA Consensus TTG ACCTTGG CHO

phocytes were used as effector cells. The addition of a CD3ε/PSMA bispecific BEAT antibody bridged the CD3ε molecule on the T lympho- cytes with the PSMA molecules on the target cell populations. This bridging activated T lymphocytes in a polyclonal manner and stimulated them to exert cytotoxic functions against the bridged target cell measured using LDH release. The non-target specific control antibody showed very low killing in all conditions. The killing mediated by the CD3/PSMA BEAT antibody, however, was correlated with the concen- tration of receptor molecules expressed. High PSMA expressing cells

described in Table 2, three cell populations were identified for each of the constructs, following a similar selection scheme as the one described for the A20 cells. Flow cytometry allowed to determine the intensity of human PSMA and GFP expression, as well as the homogeneity of the population (overlays are shown in Fig. 2, right panel).
Similar to the stable A20 cells expressing CD38, the expression level of PSMA in CHO cells was correlated with the efficiency of translation initiation of the start codon used for PSMA, confirming that this effect was not depending on the cell line used. A single cell population was found to be contaminated with a non-producing population (Cell pop- ulation 1 for start codon GTG and PSMA as GOI). This non-producing population could be easily detected with the GFP signal.
The different expression levels obtained for the three constructs were confirmed by determination of the number of specific antibodies bound per cell (sABC) for one of the CD38 expressing cell populations,

generated with the pT1-hsPSMA-ATG construct were killed in a very efficient manner, medium expressing cells (pT1-hsPSMA-GTG) were killed in an intermediate manner and low expressing cells (pT1-hsPSMA- TTG) were killed in a less efficient manner (Fig. 4).

4. Discussion
This article addresses the problem of generating homogeneous cell lines that express a target protein at low levels. Such cell lines are important for pharmacological studies, for example for safety assess- ments of monoclonal antibodies whose mechanism of action relies on the killing of target cell populations. As part of the safety assessments, it is insightful to assess whether there are conditions that allow successful killing of tumor cells (expressing medium to high levels of receptor), but not of normal body cells (expressing lower levels). Such studies may be preferably done with cell lines that naturally express the gene of interest

Fig. 2. Expression of GFP and the GOI using different start codons. Cell populations expressing CD38 in A20 cells are shown on the left panel and cell populations expressing PSMA in CHO cells are shown on the right panel. The GFP expression is shown on the x-axis and overlaid with CD38 and PSMA expression on the y-axis, respectively (Green cell population: GFP signal without APC coupled antibody; Red cell population: GFP signal and signal from APC coupled antibody specific for GOI; blue: Non-transfected control only shown for PSMA; violet: Minor population of non-expressing cells only shown for CD38). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article).

Fig. 3. sABC staining of the resulting cell lines after thawing (A) and (B) over 29 days of continuous passaging.

Fig. 4. The different levels of PSMA surface expression using the three different constructs is correlated with T-cell mediated target cell killing in presence of a CD3ε/ PSMA bispecific BEAT antibody (in blue: control using a non-specific antibody (CNT); in red: BEAT antibody; the error bars represent the standard deviation of three T-cell donors). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article).

at slightly different levels. However, such cell lines may not always be available and there may be variations of expression of the gene of in- terest based on remaining regulatory networks (e.g. cells cultivated in presence or absence of serum have different expression levels of multiple genes). In this case, using the construct presented in this article, in order to generate different levels of target protein expression in the context of a relevant cell type, may be a useful alternative.
The construct allowed us to identify stable cell populations with a standard cell line development approach, selecting for the highest GFP expressing cell populations. As the GOI is expressed from the same mRNA as GFP (with the IRES), the expression of GFP and GOI are directly correlated. This principle has been used previously to select high expressing cell lines (Freimark et al., 2010). The use of non-ATG start codons for the gene of interest reduced the translation efficiency for the GOI drastically in accordance to previous observations (Rocha et al., 1999; van Blokland et al., 2007). This decrease in translation ef- ficiency reduced the amount of GOI produced, while the unaffected expression of GFP allowed the selection of cell populations and the detection of heterogeneity of protein expression in the different cell populations.
The proposed construct is a useful tool for the generation of cell lines
with low expression levels of the transgene and in particular for the generation of cell lines with two or multiple levels of low expression of the transgene. The two examples (CD38 and PSMA) provided in the

experimental section were membrane bound proteins. However, intra- cellular or secreted proteins could be expressed in a similar way.
GFP as reporter protein can be replaced by other reporters depending on the experimental setting or availability of alternatives. The second IRES adding the resistance marker is not required and can be removed, a bicistronic construct is fully sufficient. The difference in signal between the reporter protein and the GOI could be further increased by switching the position of the GOI and the reporter (usually the first open reading frame is expressed more efficiently than the one following the IRES), modifications of the IRES sequence (and thus lowering the efficiency of ribosomal entry), or by using alternate splicing for the expression of the GOI (Lucas et al., 1996).
Combinations with other techniques may (e.g. use of unfavourable codons in the GOI, weak promoters, etc.) allow to generate an even greater variety of different expression levels of the GOI, while the fluorescent marker allows for convenient assessment of homogeneity and stability of expression.
One downside of the approach using alternate starting codons might be that the cells may not use the alternate starting codon, but rather use another ATG codon coding for methionine in the open reading frame of GOI to initiate translation. In this example, two genes coding for human CD38 and for human PSMA were successfully expressed, although they have 4 and 15 internal ATG codons (coding for methionines) respec- tively. This indicates that at least for these two molecules, additional

internal ATG codons may not be used efficiently as starting codons or may not lead to functional product.
Inducible systems might be used for the same purpose. However, they are limited due to the necessity of the presence of the co-activator at different concentrations in order to obtain different levels of expression. This might not be feasible in in vitro or in vivo experiments and therefore represents an important limitation of the inducible system. Standard cell line development approaches based on random integration would be limited by the presence of potentially unstable and heterogeneous cell populations that would be difficult to distinguish from low expressing homogeneous cell populations. Targeted integration approaches may partially overcome this limitation, but are technically much more challenging than the approach feasible with the construct described in this paper.
In summary, we believe that the proposed construct allows a
convenient way to select homogeneous cell populations expressing low amounts of proteins of interest following a convenient and simple random-integration approach.
CRediT authorship contribution statement
Dominique Schlicht: Conceptualization, Data curation, Methodol- ogy. Carole Estoppey: Data curation, Methodology. Julie Macoin: Data curation, Methodology. Blandine Pouleau: Data curation, Methodol- ogy. Martin Bertschinger: Conceptualization, Methodology, Writing – original draft, Writing – review & editing.
Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Acknowledgments
The authors would like to thank Dr Dean Thomas, Dr Benjamin Moritz and Dr Christel Aebischer for helpful discussions and critical review of the article.

Appendix A. Supplementary data
Supplementary material related to this article can be found, in the online version, at doi:https://doi.org/10.1016/j.jbiotec.2020.10.008.
References
DeMaria, C.T., Cairns, V., Schwarz, C., Zhang, J., Guerin, M., Zuena, E., Estes, S., Karey, K.P., 2007. Accelerated clone selection for recombinant CHO CELLS using a FACS-based high-throughput screen. Biotechnol. Prog. 23 (2), 465–472.
Fallot, S., Ben Naya, R., Hieblot, C., Mondon, P., Lacazette, E., Bouayadi, K., Kharrat, A., Touriol, C., Prats, H., 2009. Alternative-splicing-based bicistronic vectors for ratio- controlled protein expression and application to recombinant antibody production. Nucleic Acids Res. 37 (20), e134.
Freimark, D., Jerome, V., Freitag, R., 2010. A GFP-based method facilitates clonal selection of transfected CHO cells. Biotechnol. J. 5 (1), 24–31.
Kaufman, R.J., 2000. Overview of vector design for mammalian gene expression. Mol.
Biotechnol. 16 (2), 151–160.
Lucas, B.K., Giere, L.M., DeMarco, R.A., Shen, A., Chisholm, V., Crowley, C.W., 1996. High-level production of recombinant proteins in CHO cells using a dicistronic DHFR intron expression vector. Nucleic Acids Res. 24 (9), 1774–1779.
Mancia, F., Patel, S.D., Rajala, M.W., Scherer, P.E., Nemes, A., Schieren, I., Hendrickson, W.A., Shapiro, L., 2004. Optimization of protein production in mammalian cells with a coexpressed fluorescent marker. Structure 12 (8), 1355–1360.
Qin, J.Y., Zhang, L., Clift, K.L., Hulur, I., Xiang, A.P., Ren, B.Z., Lahn, B.T., 2010. Systematic comparison of constitutive promoters and the doxycycline-inducible promoter. PLoS One 5 (5), e10611.
Quax, T.E., Claassens, N.J., Soll, D., van der Oost, J., 2015. Codon bias as a means to fine- tune gene expression. Mol. Cell 59 (2), 149–161.
Rocha, E.P., Danchin, A., Viari, A., 1999. Translation in Bacillus subtilis: roles and trends of initiation and termination, insights from a genome analysis. Nucleic Acids Res. 27 (17), 3567–3576.
Shi, L., Chen, X., Tang, W., Li, Z., Liu, J., Gao, F., Sang, J., 2014. ’Combination of FACS and homologous recombination for the generation of stable and high-expression engineered cell lines’. PLoS One 9 (3), e91712.
van Blokland, H.J., Kwaks, T.H., Sewalt, R.G., Verhees, J.A., Klaren, V.N., Siersma, T.K.,
Korse, J.W., Teunissen, N.C., Botschuijver, S., van Mer, C., Man, S.Y., Otte, A.P., 2007. A novel, high stringency selection system allows screening of few clones for high protein expression. J. Biotechnol. 128 (2), 237–245.
Zhu, M., Wu, B., Brandl, C., Johnson, J., Wolf, A., Chow, A., Doshi, S., 2016. ’Blinatumomab, a bispecific T-cell engager (BiTE((R))) for CD-19 targeted cancer immunotherapy: clinical pharmacology and its implications’. Clin. Pharmacokinet. 55 (10), 1271–1288.CL13900