Reducing recombinant protein expression during CHO pool selection enhances frequency of high-producing cells

Journal of Biotechnology 296 (2019) 32–41

Contents lists available at ScienceDirect

Journal of Biotechnology journal homepage: www.elsevier.com/locate/jbiotec

Reducing recombinant protein expression during CHO pool selection enhances frequency of high-producing cells Adeline Poulaina,b, Alaka Mullicka, Bernard Massiea,b, Yves Durochera,c,

T

⁎

a

National Research Council Canada, 6100 Royalmount Avenue, Montréal, QC H4P 2R2, Canada Département de microbiologie et immunologie, Faculté de médecine, Université de Montréal, QC, Canada c Département de biochimie et médecine moléculaire, Faculté de médecine, Université de Montréal, QC, Canada b

A R T I C LE I N FO

A B S T R A C T

Keywords: CHO cells Recombinant protein production Inducible expression Cumate gene-switch BiPmethionine sulfoximine

Chinese hamster ovary (CHO) cells are the most widely used mammalian host for industrial-scale production of monoclonal antibodies (mAbs) and other protein biologics. Isolation of rare high-producing CHO cell lines from heterogeneous populations of stable transfectants is a daunting task and delays the process of manufacturing of novel biologics. A variety of factors that contribute to the low frequency of high-producing clones have been described; however, the impact of metabolic burden and other stresses (eg. ER stress) associated with sustained high-level expression of recombinant protein (r-protein) during selection of stable transfectants has not been fully appreciated. CHO cell line development has not traditionally received much optimization in this area because the vast majority of platforms use constitutive expression systems to produce biologics. Previously, we developed a cell line (CHOBRI/rcTA) containing a robust inducible expression system, based on the cumate gene switch, that allows r-protein expression to be down-regulated during selection. Using this switch, we generated inducible CHOBRI/rcTA pools expressing an Fc-fusion protein within two weeks of transfection with volumetric productivity of up to 1.1 g/L at 17 days post-induction in a fed-batch culture process. Herein, we show that the ability to regulate r-protein expression during pool generation confers a substantial advantage for selecting highproducing stable clones. Reducing expression levels (“off-state”) during pool selection dramatically enhances high-producer frequency compared to a pool in which expression was maintained at a high level during selection (“on-state”, mimicking a constitutive expression system). Overexpression of the r-protein during the pool selection process negatively affects pool recovery and is associated with subtle but significant increases in BiP expression and cell death compared to pool selection in the “off-state”. Our data shows that the cumate gene switch is a valuable platform for stable clone generation and supports the wider application of inducible systems for scalable production of biologics in CHO cells.

1. Introduction Chinese hamster ovary (CHO) cells are the most widely used mammalian host for industrial-scale production of monoclonal antibodies (mAbs) and other protein biologics (Durocher and Butler, 2009). Many factors, such as their ability to carry out human-like post-translational modifications (Butler and Spearman, 2014; Lalonde and Durocher, 2017), their resistance to infectious human viruses (Berting et al., 2010), their ease of growth in chemically defined serum-free medium in large scale cultures (Kim et al., 2012) and their lengthy track record for manufacturing safe and efficient biologics (Wurm, 2004) explain this popularity. However, despite many advantages, the development of a robust stable cell line, producing high levels of

recombinant protein, is a labour-intensive process that typically requires 6–9 months. Cell line generation begins with transfection of host cells with a plasmid vector containing both the gene of interest and a selection marker, followed by selection of cells having integrated the plasmid vector into their genome. The selection process generates a heterogeneous pool of clones with a variety of chromosomal integration sites containing different numbers of integrated transgenes. Stable pools show high diversity in terms of cell-specific expression level and growth (Lai et al., 2013), which is in part attributed to the site of integration and copy number of the transgene and in part due to the intrinsic heterogeneity existing in the parental cell line (Pilbrough et al., 2009). Because cells that combine desired phenotypes for growth, productivity and stability are rare, isolation of a high-producing clone

⁎

Corresponding author at: National Research Council Canada, 6100 Royalmount Avenue, Montreal, QC H4P 2R2, Canada. E-mail addresses: [email protected] (A. Poulain), [email protected] (A. Mullick), [email protected] (B. Massie), [email protected] (Y. Durocher). https://doi.org/10.1016/j.jbiotec.2019.03.009 Received 30 October 2018; Received in revised form 12 March 2019; Accepted 14 March 2019 Available online 15 March 2019 0168-1656/ © 2019 Published by Elsevier B.V.

Journal of Biotechnology 296 (2019) 32–41

A. Poulain, et al.

sulfoximine (MSX). In the current study, we examined if the ability to control r-protein expression during selection could be advantageous not only for generating stable CHOBRI/rcTA pools but also for the isolation of high-producing cell lines. Using the Fc-fusion protein hCD200Fc as a model protein, we have demonstrated that reduced levels (“off-mode”) of expression during MSX selection leads to the generation of stable CHOBRI/rcTA pools with 2-fold increased productivity in a fed-batch culture process compared to pools in which high level (“on-mode”) expression was induced by cumate addition. Sub-population distribution study, performed by the isolation of μ-pools (obtained by plating 5 cells/well), revealed a significant higher frequency of cells with high specific productivity in the pool selected in the absence of cumate. Moreover, we observed that maintaining a high level of hCD200Fc expression upregulated ER-resident glucose regulated proteins GRP78/ Bip concomitant with a significant decrease in viable cell number and proliferation during the early phase of the selection process, compared to non-induced pool. Our data suggest that during MSX pool selection, some cells constitutively expressing high level of r-protein and with a higher level of ER stress will not survive or will be overgrown by lowexpressing cells, more capable to cope with the stresses. Thus, the use of an inducible system allowing the reduction of r-protein production during the pool selection phase could offer an advantage in the isolation of high-producing stable clones by increasing their frequency in the pool, thus facilitating clone screening and reducing costs.

from a heterogeneous population represents a major endeavor since it requires the screening of hundreds to thousands of individual clones. In addition to integration site, copy number, and host cell heterogeneity, there are other factors that likely contribute to the low frequency of high-producing clones, including a combination of cellular stresses triggered by the transfection process itself, the overexpression of the gene of interest, and the selection process that often uses cytotoxic drugs. It is likely that high producing clones are overgrown by the faster-growing low- or non-producing ones as a result of these stresses (Kromenaker and Srienc, 1994; Lee et al., 1991). High levels of r-protein in the endoplasmic reticulum (ER) can activate signalling cascades known as the unfolded protein response (UPR), endoplasmic reticulum overload response (EOR) and ER-associated protein degradation (ERAD) to restore ER homeostasis (Johari et al., 2015; Ku et al., 2010; Zhou et al., 2018). In a first phase, activation of UPR promotes cellular survival by reducing unfolded protein load through several pro-survival mechanisms, such as attenuation of protein synthesis, upregulation key components of the protein folding machinery or through cell cycle arrest in G1 phase (Hetz and Papa, 2018). When ER stress is prolonged and homeostasis is not restored, the UPR can trigger apoptosis (Jager et al., 2012; Moore and Hollien, 2012; Zhou et al., 2018). Thus, cells constitutively expressing high levels of r-protein may consequently suffer from ER stress and be less fit to grow or survive during the pool selection process. This phenomenon could also be exacerbated by the presence of the selection drug, which may further sensitise cells to apoptosis or necrosis. It would theoretically be beneficial to limit r-protein expression, using an inducible expression system, during the selection phase of CHO cell line development in order to reduce cellular stresses that could limit growth or survival of potential high-producing clones. However, so far, the evaluation of inducible systems in this context has been limited. In a previous study (Hu et al., 2013), it was speculated that during stable clone selection for the expression of “difficult-toexpress’’ antibodies, high-expressing cells not able to cope with ER stress would die. To circumvent this, the use of inducible gene expression systems, allowing the regulation of transgene expression during the cell line development process, could be a solution. Accordingly, the use of a tetracycline responsive promoter to produce difficultto-express mAbs in stable CHO cell lines was shown to reduce clonal instability, likely arising from intracellular accumulation of misfolded antibodies over time and after cryopreservation (Misaghi et al., 2014). Using the same inducible expression system, stable clones were successfully developed and scaled-up in bioreactors to produce moderate levels of an engineered version of human DNaseI, known to be cytotoxic when overexpressed in CHO cells (Lam et al., 2017). Although these studies clearly point to the advantages of using an inducible system when dealing with “cytotoxic” targets, its usefulness for overexpressing r-proteins that do not a priori affect cell growth or viability when overexpressed is unclear. Therefore, we wanted to examine whether the use of an inducible system for r-protein production could be more broadly advantageous, both for generating pools and subsequently isolating clones with higher productivities. We previously developed a cell line for the inducible expression of rproteins based on the cumate gene switch. Using viral vectors, we have shown that the cumate-regulated promoter is not only inducible but also produces higher yields of various r-proteins than the constitutive human CMV promoter in CHO cells (Gaillet et al., 2007; Gilbert et al., 2007). Furthermore, we have recently shown that the cumate-inducible expression system, which combines the CHOBRI/rcTA cell line and plasmids containing the cumate-responsive CR5 promoter, was generation pools more productive than using constitutive promoters and allowed for the rapid generation of stable CHO pools (< 3 weeks) producing high levels of r-proteins (400–900 mg/L) (Poulain et al., 2017). Moreover, we showed that CHO pool volumetric productivities were not affected by freeze-thaw cycle or following maintenance in culture for over one month in the presence of the selection drug methionine

2. Materials and methods 2.1. CHOBRI/rcTA cell line culture condition CHOBRI/rcTA cell line has been described in detail in our previous study (Poulain et al., 2017). Briefly, this cell line contains an inducible expression system based on the reverse activator configuration of the cumate-gene switch (Mullick et al., 2006). CHOBRI/rcTA cells stably express both the cymene repressor (CymR), able to bind to the CuO operator sequence in the absence of cumate, and the cumate reverse transactivator (rcTA) composed of the VP16 activation domain fused to the reverse CymR, able to bind to the CuO sequence when the CymR is released from it in the presence of cumate. Upon transient or stable transfection of a plasmid vector encoding a gene of interest driven by the CR5 inducible promoter (which contains five CuO operator sequences), expression is activated by the addition of cumate (p-isopropylbenzoate), a small non-toxic inducer. Cells were grown in PowerCHO2 chemically defined medium (Lonza, Walkersville, MD, USA) supplemented with 4 mM L-glutamine. The cells were grown at 37 °C and 5% CO2 under constant agitation (120 rpm) and passaged three times per week at cell density of 0.1 × 106 to 0.2 × 106 cells/mL in 125 mL or 250 mL Erlenmeyer flasks (Corning, NY, USA), with 25 or 50 mL culture medium, respectively. Cell number and viability were measured with the Cedex Innovatis automated cell counter Cedex Analyzer (Roche, Laval, Qc) using the trypan blue exclusion method. 2.2. Cumate Cumate was purchased from Ark Pharm Inc (Arlington Heights, IL, USA) and dissolved at 2 mg/ml in 95% (v/v) ethanol. The cumate solution was aliquoted in sterile screwcap Eppendorf tubes and stored at 2–8 °C for up to 16 months. 2.3. Plasmid vectors The pTT75® plasmid vector used in this study was derived from the pTT® vector backbone (Durocher et al., 2002) in which the CMV5 promoter was replaced with CR5 promoter, a short sequence from the scaffold attachment region 3 (SAR3) of the human interferon α2 gene and the human glutamine synthase (GS) gene driven by the SV40 33

Journal of Biotechnology 296 (2019) 32–41

A. Poulain, et al.

was done at 280 nm. Quantification was achieved by integrating the area under the elution peak and comparing it against an IgG standard curve.

promoter were inserted. The hCD200Fc (Gaillet et al., 2007, 2010) is a fusion protein consisting of the human CD200 extracellular domain linked to a human IgG1 Fc domain. The cumate inducible CR5 promoter was described previously (Mullick et al., 2006). The hCD200Fc gene was removed from the pTT75-hCD200Fc plasmid by restriction enzyme digestion to generate the empty pTT75 vector. All plasmids were amplified in Escherichia coli (DH5α) and purified using an in-house lowendotoxin chromatographic method (unpublished).

2.8. Generation of μ-pools Cells from Cum0 and Cum1 selected pools were plated in 384-clear bottom plates (Corning #3712) at a concentration of 5 cells/well in 60 μl of BalanCD Growth A medium and incubated at 37 °C and 5% CO2 for 10 days until 30–50% confluence was obtained in most wells. Then, 20 μl of BalanCD Growth A medium supplemented with 2 μg/ml of cumate was added in each well to obtain a final cumate concentration of 0.5 μg/ml. Plates were incubated at 37 °C under humidified atmosphere containing 5% CO2 for 5 days. On day 5 post-induction, cell confluence was determined, 30 μl of supernatant was collected in each well, and hCD200Fc titer was determined by Homogeneous Time Resolved Fluorescence (HTRF) using the human IgG kit (Cisbio, France), according to manufacturer’s recommendations. To determine the % confluence in each well, plates were imaged using the ImageXpress Micro XLS wide field high-content analysis system (Molecular Devices, CA, USA). Then, the % confluence was calculated by dividing the surface area occupied by the cells by the total well surface area. Specific growth rate during production phase was calculated using the following formula, where CD5 corresponded to the % confluence at day 5 post-induction and CDind the % confluence at the day of induction:

2.4. pDNA transfection Cells were transfected using linear polyethylenimine (PEIpro™, Polyplus-Transfection, Illkirch, France) as previously described (Poulain et al., 2017). 2.5. MSX selection of CHOBRI/rcTA cell pools CHOBRI/rcTA cell pools were generated by transfecting cells with the expression vectors as described above. The day after transfection, the cells were centrifuged for 5 min at 250 rpm and seeded at density of 0.5 × 106 cells/mL in selection medium (PowerCHO2 medium supplemented with 25 μM of methionine sulfoximine) without (Cum0 pool) or with 1 μg/mL of cumate (Cum1 pool). Selection medium (supplemented or not with 1 μg/mL of cumate) was replaced every 2–3 days during 14–18 days while maintaining viable cell density at 0.5 × 106/ mL. Cell number and viability were measured with the Cedex Innovatis automated cell counter Cedex Analyzer as described above. When cell viability reached greater than 95%, pools were inoculated at 0.2 × 106 cells/mL in 125 or 250 mL Erlenmeyer flasks.

μ (day−1) = Ln (CD5 – CDind)/Δt

2.6. Intracellular hCD200Fc labelling

2.9. Apoptosis and necrosis determination

During CHO pool selection, 1 × 106 cells were collected every 2–3 days, washed in PBS, fixed with 1 mL ice-cold 70% ethanol and stored at −20 °C prior to staining. The thawed cell pellets were washed twice with PBS, resuspended in 1 mL of PBS containing CloneDetect antiHuman IgG (Fc) specific antibody conjugated to fluorescein isothiocyanate (FITC; Molecular Devices, Sunnyvale, CA) and incubated in the dark under continuous agitation for 1 h at room temperature. Cell pellets were washed, resuspended in 500 μL of PBS and filtered through a 50 μm mesh Nitex tissue (Sefar Group, Heiden, CH). Flow cytometry analysis was done with a BD LSR Fortessa™ cell analyzer (BD Biosciences, Mississauga, ON, Canada). Each sample was measured with a minimum of 10,000 recorded events.

Apoptosis and necrosis were determined using the Annexin V-propidium iodide assay and flow cytometry according to the manufacturer’s instructions (Annexin V-FITC Apoptosis Detection Kit, Sigma). Briefly, the harvested cells were washed twice with DPBS and resuspended in binding buffer at concentration of 1 × 106 cells/mL and incubated for 10 min at 37 °C with Annexin V-FITC. Propidium iodide (PI) (1 μg/mL) was added immediately before flow cytometry analysis. Fifty thousand cells were recorded in each analysis. The population of cells with no staining by either PI or Annexin V-FITC conjugate was identified as live cells. Cells stained with the Annexin V-FITC conjugate alone were identified as early apoptotic cells, and cells that were labeled by both propidium iodide and Annexin V-FITC conjugate were identified as late-apoptotic or necrotic cells.

2.7. Fed-batch culture production 2.10. Intracellular hCD200Fc and Bip expression analysis by western blotting

CHOBRI/rcTA cell pools were inoculated in 125 mL Erlenmeyer flask at 0.3 × 106 cells/mL in 20 mL of BalanCD Growth A chemically defined medium (Irvine Scientific, Irvine, CA, USA) supplemented with 25 μM of MSX and incubated at 37 °C and 5% CO2 under constant agitation (120 rpm). At day 3 post-inoculation, when cell density reached 2.5–3.5 × 106 cells/mL, expression of the recombinant protein was induced by adding 2 μg/mL of cumate. MSX concentration was adjusted to 125 μM, and F12.7 feed (Irvine Scientific) was added followed by a temperature shift to 32 °C. Every 2–3 days, cultures were fed with 5% (v:v) F12.7 and samples were collected for recombinant protein (pA-HPLC) and glucose (VITROS 350, Orthoclinical Diagnostics, USA) concentration determination. Glucose was added to maintain a minimal concentration of 17 mM. Protein A high performance liquid chromatography (WATERS Corporation, Milford, MA) used a 2.1 mmD × 30 mmH, 104 μL, POROS®A20 column (Invitrogen, Grand Island, NY). Before their injection on the column at a flow rate of 2 mL/ min, samples were filtered by centrifugation at 8,000–11,000 × g for 3 min through a NANOSEP MF GHP 0.45 μm filter (PALL Life Sciences). Elution was performed using 0.15 M NaCl, pH 2.0, and UV detection

For intracellular protein extraction, 1 × 106 cells were harvested, washed twice with cold PBS and lysed in 100 μL of RIPA buffer (50 mM Tris–HCl pH 8, 150 mM NaCl, 1% (v/v) Nonidet P-40, 0.5% (w/v) sodium deoxycholate, 0.1% (w/v) SDS), supplemented with protease (cOmplete™, Roche) and phosphatase (PhosSTOP™, Roche) inhibitor cocktails and incubated on ice for 10 min. Lysates were centrifuged for 10 min at 12,000 rpm and clarified extracts harvested and stored at −80 °C. SDS-PAGE was performed under reducing (50 mM DTT) conditions using Criterion™ 4–12% Bis-Tris Protein Gel (Biorad) with 2-(Nmorpholino)ethanesulfonic acid (MES) running buffer. Proteins were transferred to a 0.2 μm nitrocellulose membrane for 7 min using TransBlot® Turbo™ RTA Kit (Bio-Rad, cat# 170-4271) and Trans-Blot® Turbo™ Transfer System (Bio-Rad, cat# 170-4150). Before immunodetection, the membranes were stained with 0.2% w/v Ponceau S. Membranes were then blocked with 5% (w/v) BSA in TBS-T and then probed with anti-BiP primary antibody (1:4000; Cell Signaling Technology) followed by detection using an HRP-conjugated anti-rabbit 34

Journal of Biotechnology 296 (2019) 32–41

A. Poulain, et al.

hCD200Fc compared to Cum0 cells (black bars) throughout the selection phase. To rule out the possibility of a toxic effect of cumate itself or of the rcTA transactivator (whose expression is induced by cumate addition), a pool transfected with the empty pTT75 plasmid was also selected without or with 1 μg/ml of cumate (Fig. 1B). No difference was observed between the two pools, and interestingly, viability did not fall below 80% during MSX selection. The cell growth rate measured during the first days of selection was lower (or even negative at Days 4 and 6) for the Cum1 pool expressing hCD200Fc compared to the Cum0 pool (Fig. 1C). Clearly, cumate had no effect on the pool transfected with empty pTT75 plasmid since growth rates were the same for both Cum0 and Cum1 pools (Fig. 1D). These observations suggest that hCD200Fc overexpression per se negatively impacts viability and cell growth in the early days of MSX selection.

IgG or probed with HRP-conjugated anti-human IgG Fc fragment (1: 5000; Sigma). Pictures were recorded with a Chemidoc MP Imaging System (Bio-Rad). 3. Results 3.1. Recombinant protein overexpression during selection negatively affects cell growth and viability The pTT75-hCD200Fc plasmid, carrying the hCD200Fc gene under the control of the CR5 inducible promoter, was transfected into the CHOBRI/rcTA cell line. The day following transfection, MSX selection was applied and the cells were split into two distinct populations, one left without cumate addition (Cum0 pool) and the other one where hCD200Fc expression was induced by addition of 1 μg/ml of cumate (Cum1 pool). Pool viability was determined every two days by trypan blue exclusion and steady-state intracellular hCD200Fc protein was monitored at various days during the selection process by labelling with an FITC-conjugated anti-human IgG (Fc) specific antibody. Labelled cells were analyzed by flow cytometry and the fluorescence index ([FITC-fluorescence mean] x [percent FITC-positive cells]) of each pool was determined. As expected, addition of 1 μg/ml of cumate induced hCD200Fc expression (Fig. 1A). A basal level of hCD200Fc expression was observed for the Cum0 pool as a result of the leakiness of the cumate expression system as previously described (Gaillet et al., 2007; Poulain et al., 2017). Interestingly, the Cum1 pool experienced a more pronounced decline in cell viability that dropped to 56% at Day 6, a level significantly lower than with the Cum0 pool (> 70%). However, from Day 11, both pools recovered and showed the same viability (Fig. 1A). Flow cytometry analyses confirmed that Cum1 cells (empty bars) contained 2- to 5-fold higher steady state levels of intracellular

3.2. Recombinant protein overexpression during pool selection increases Bip expression and annexin V-positive cells Cells amenable to high-level expression of r-protein may experience a variety of stress conditions, particularly endoplasmic reticulum (ER)associated stress (Johari et al., 2015). The ER-resident Glucose Regulated Protein 78 (GRP78/Bip) is a protein chaperone that is involved in the normal process of protein folding during biosynthesis but also binds and retains misfolded proteins within the ER lumen; increased Bip expression is a marker for the onset of ER stress (Hendershot, 2004; Ku et al., 2010; Le Fourn et al., 2014). To evaluate ER stress in Cum0 and Cum1 pools during the selection process, Bip expression level was examined by western blot. To discriminate stress caused by r-protein overexpression from stress due to MSX selection, Bip level was also monitored in cells transfected with empty pTT75 plasmid. Intracellular hCD200Fc and Bip level in pTT75-hCD200Fc and pTT75 cell lysates

Fig. 1. hCD200Fc expression during pool selection negatively impacts cell growth and viability. CHOBRI/rcTA cells were transfected with pTT75-hCD200Fc (panels A,C,D) or empty pTT75 (panels B,D,F) vectors. Twenty four hours later, cells were split in two flasks and selected with 25 μM of MSX in absence (Cum0 pool) or presence (Cum1 pool) of 1 μg/mL of cumate. Cell viability (Cedex) and intracellular hCD200Fc expression level (FITC-labelled anti-Human IgG; flow cytometry) were monitored at various times during the selection period. The fluorescence index is calculated by multiplying the [FITC-fluorescence mean] by the [% FITC-positive cells]. Specific growth rate (C and D) were calculated at various time points for each pool. Data shown are the mean values and standard deviations of two independent biological replicates. 35

Journal of Biotechnology 296 (2019) 32–41

A. Poulain, et al.

Fig. 2. hCD200Fc expression during pool selection increases Bip expression and annexinV-positive cells. Panel A: CHOBRI/rcTA cells were transfected with pTT75-hCD200Fc or empty pTT75 vector and cells were split in two flasks 24 h later (Day 0) and selected with 25 μM of MSX in absence (Cum0 pool) or in presence (Cum1 pool) of 1 μg/mL of cumate. Cells were sampled every two days during the first ten days of the selection for further analyses. Bip (78 kDa) and hCD200Fc (47 kDa under reducing conditions) intracellular expression was analyzed by western-blot from the same nitrocellulose membrane. The membrane was also stained with Ponceau Red to monitor for protein load (the asterisks point to the position of hCD200Fc). Panel B: intracellular hCD200Fc expression level was reported as fold-induction over Day 0 (Cum0) after normalization for protein load. Panel C: Bip expression level was reported as foldinduction relative to its level measured at Day 0 (just prior to MSX addition) and were normalized for protein load. The % of (AnnexinV+/PI-) cells representing early apoptotic cells (panel D) and the % of (Annexin+/PI+) cells representing necrotic and late apoptotic cells (panel E) were determined for each pool. Data shown are the mean values and standard deviations of two independent biological replicates (except for panel A which shows results from a single representative experiment). Statistical significance was determined to be P < 0.01 (*) using the Holm-Sidak method.

of hCD200Fc expression was also observed in the absence of cumate, especially at 24 h post-transfection (Day 0), were the plasmid copy number per cells is expected to be maximal (Carpentier et al., 2007). Bip expression increased significantly after 4 days of MSX selection for both pTT75-hCD200Fc and empty pTT75 pools, possibly due to a stress induced by intracellular glutamine depletion (Fig. 2A and C). Although the impact of low level of r-protein expression on Bip level in noninduced cells was difficult to discriminate from the one caused by MSX

were all analyzed on the same nitrocellulose membrane (Fig. 2A). As expected and in agreement with Fig. 1A, the addition of cumate in the culture transfected with pTT75-hCD200Fc increased intracellular hCD200Fc steady-state level (Fig. 2B). It is noteworthy that the hCD200Fc protein was not detectable by Ponceau staining of the membrane (position on the membrane denoted by an asterisk), suggesting that the intracellular hCD200Fc steady-state level is fairly low, even in the cumate-induced state. As previously observed, a basal level 36

Journal of Biotechnology 296 (2019) 32–41

A. Poulain, et al.

cells/well in order to obtain a large number of oligo-clonal populations, referred as micro-pools (μ-pools). After plating, cells were left to grow without MSX for 10 days in the absence of cumate, until 30–50% of confluence (measured using the ImageXpress) was obtained in most wells (growth phase). Then, cumate was added in each well and hCD200Fc titers were quantified five days later (production phase) using a HTRF-assay. Plates were also imaged on the ImageXpress microscope to measure confluence for each μ-pool at the end of the 5 days production (Fig. 4A). By seeding 5 cells per well, a μ-pool outgrowth efficiency close to 100% was obtained since cell growth (> 10% confluence) was observed in 1146 and 1139 wells out of the 1152 seeded wells for both Cum0 and Cum1 pools, respectively. Gaussian distribution histograms showed a large heterogeneity regarding the % confluence at the time of induction, but this heterogeneity was similar for both Cum0 and Cum1 μ-pools (Fig. 4B). The frequency of low-confluence wells (≤40%) was slightly higher for Cum1 μ-pools compared to Cum0 μ-pools. At the end of the production phase, the % confluence distribution was similar for both pools, suggesting that proliferation during production phase was similar (Fig. 4C). This indicates that induced Cum0 μ-pools had slightly lower specific growth rate compared to induced Cum1 μ-pool, as shown in Fig. 4D, and as observed previously for the pools in fed-batch (Fig. 3D). Remarkably, the hCD200Fc titers of μ-pools derived from Cum0 pool were higher than those derived from Cum1 pool and, interestingly, only one μ-pool producing more than 100 mg/L was observed in the Cum1 pool, while 14 were found in Cum0 pool (Fig. 5A). For each μ-pool, a relative specific productivity was determined by normalizing hCD200Fc titer with the % confluence measured at the end of the production phase. Cum0 μ-pools had higher specific productivity compared to Cum1 μ-pools (Fig. 5B). The average specific productivity for μ-pools isolated from the Cum0 pool was twice that of those isolated from Cum1 pool, confirming the data obtained for the pools themselves (Fig. 3F). In addition, the Cum0 pool was clearly enriched in high producing μ-pools since 22% of them had a relative specific productivity greater than 40 while only 3.3% were found in the Cum1 pool (Fig. 5C). Interestingly, μ-pools with low productivity (< 10) were much less abundant in the Cum0 pool. In summary, we observed that the Cum0 pool was enriched with cells having higher specific productivity.

selection itself, a slight upregulation of Bip, independent of MSX, was observed in cells overexpressing hCD200Fc (compare Cum1 to Cum0 in panels A and C). To further describe the type of cell death as initially observed with trypan blue exclusion, annexinV binding and PI uptake were used to measure apoptosis and necrosis occurring during the selection process. The percentage of early apoptotic (annexinV-positive/ PI-negative) cells is represented in Fig. 2D, while the percentage of late apoptotic/necrotic cells (annexinV-positive/PI-positive) is represented in Fig. 2E. No difference was observed between pools for cells in early apoptosis (Fig. 2D). However, cells transfected with pTT75-hCD200Fc and overexpressing hCD200Fc (Cum1 pool) showed a significantly higher percentage of late apoptotic/necrotic cells compared to Cum0 pool, or to pools expressing no r-protein (pTT75 pools) (Fig. 2E). These results suggest that concomitant MSX selection and high level hCD200Fc expression further exacerbates cellular stresses that could ultimately lead to cell death. 3.3. Reducing r-protein expression during selection increases pool productivity To evaluate whether reducing r-protein expression during selection affects pool performance, pools that recovered from MSX selection under induced and non-induced conditions were compared using a fedbatch culture process. For both pools, a fed-batch culture process without cumate induction was also performed (non-induced pool). Viability and viable cell density were determined until Day 17 when cell viability dropped to ˜60% (Fig. 3A). No viability difference was observed during the fed-batch process for both Cum0 and Cum1 pools in the induced (+ cumate) or non-induced (- cumate) state. However, hCD200Fc overexpression (+ cumate) reduced viable cell density by 38% for Cum0 pool, in comparison to the non-induced (- cumate) pool (Fig. 3B). Indeed, the maximal viable cell density for the induced culture was around 1.0 × 107 cells/mL compared to 1.6 × 107 cells/mL for the non-induced culture, resulting in a 30% reduction in the integral of viable cell density (IVCC) calculated at the end of the fed batch culture (Fig. 3C). As for the Cum1 pool, only a slightly more pronounced drop in viable cell density was observed from day 12 in the presence of cumate (Fig. 3B) while no difference was observed between induced and non-induced cultures for the IVCC (Fig. 3C). Moreover, during the exponential growth phase, specific growth rates were the same for the induced and non-induced Cum1 pool, while a significant decrease was observed for the Cum0 pool in the induced state (Fig. 3D). Interestingly, the specific growth rate of the Cum0 pool in the induced state was similar to Cum1 pool in the non-induced or induced state. To compare volumetric productivities, samples were collected every 2–3 days and hCD200Fc titers quantified by protein-A HPLC (Fig. 3E). The average titer for Cum0 pool at day 17 (1.1 g/L) was around 2-fold higher compared to the Cum1 pool (0.6 g/L). The induced/non-induced titer ratios for both Cum0 and Cum1 pools were around 12. Specific productivity was determined from the volumetric titers and IVCC every 2–3 days (Fig. 3F). A two-fold higher specific productivity was calculated for the induced Cum0 pool (peak at 14 pg/cell/day) compared to the induced Cum1 pool (peak at 7 pg/cell/day) for the whole duration of the fed-batch culture. Because induced Cum0 and Cum1 pools had similar IVCCs, the difference observed in the final fed-batch titers could be explained by an enrichment of the Cum0 pool with cells with higher specific productivity.

4. Discussion In industrial CHO cell line development, the main objective is to select a clone capable of high and stable product expression level whilst maintaining rapid cell growth and high viability. Despite significant efforts to develop new technology platforms to shorten the time of biologics development, isolation of cell lines meeting manufacturing process criteria is still a major challenge, requiring the screening of several hundreds to thousands of individual clones from a heterogeneous population of transfectants. The low occurrence of high producing clones is likely the result of a combination of various constraints including the integration site, integrated copy number and cellular stresses caused by the transfection process, selection agent and overexpression of the gene of interest. Some of these effects can also be protein-specific (Sommeregger et al., 2016): for instance, it is not surprising that high-level expression of cytotoxic/cytostatic r-proteins is not well tolerated by the host cell. In these cases, reducing expression of the gene of interest during selection is beneficial. Previously, the cumate switch was used to generate stable transfectants that can be induced to express vesicular stomatitis virus glycoprotein (VSVg) and granulocyte colony stimulating factor (GCSF), two proteins whose constitutive expression causes a dramatic decrease in cell viability ((Broussau et al., 2008) and our unpublished results). Other inducible systems, such as a tetracycline-inducible system, have similarly been used to produce “difficult-to-express” proteins (Lam et al., 2017; Misaghi et al., 2014). However, it was not clear what advantage an inducible system may have for the expression of a protein that does not

3.4. Reducing r-protein expression during pool selection enhances frequency of high producers To compare the frequency of high-producer cells in the pools, cells from each pool were distributed in 384-well plates under quasi-limiting dilution. Given the low efficiency of single cell cloning (< 60% of cell outgrowth following 1 cell/well plating, data not shown) and the need to analyse many cells for statistical significance, wells were seeded at 5 37

Journal of Biotechnology 296 (2019) 32–41

A. Poulain, et al.

Fig. 3. Performance of Cum0 and Cum1 pools during a fed-batch culture process. Viability (A) and viable cell density (B) were monitored during fed-batch cultures for both induced and non-induced Cum0 and Cum1 pools. The integral of viable cell concentration (IVCC) was calculated over the 17 days fed-batch duration (panel C). Specific growth rate was calculated during the exponential phase between day 2 and day 7 (panel D). Volumetric productivity of induced and non-induced pools was evaluated at the end of the 17 days fed-batch cultures (panel E) and specific productivities were determined from titers and corresponding IVCC (panel F). hCD200Fc titers were determined by pA HPLC. Data shown are the mean values and standard deviations of two independent biological replicates. Statistical significance was determined to be of P < 0.05 (*) and P < 0.01 (**).

same conditions, in the presence of cumate. We found that repressing hCD200Fc expression during MSX selection increased pool productivity by ˜two-fold at the end of the fed-batch process (1.1 vs 0.6 g/L). The increased volumetric productivity was mostly due to a two-fold higher specific productivity. While cumate induction of the Cum0 pool significantly reduced growth rate and IVCC during the fed-batch process, it had no impact on Cum1 pool, suggesting that the latter had adapted to r-protein synthesis during selection. Moreover, for both pools, overexpression of hCD200Fc (induced state) did not affect cell viability during the fed-batch process. Thus, although hCD200Fc was not toxic to

have overt cytotoxic/cytostatic effects. In this study we have explored the impact of reducing expression of hCD200Fc during the selection process on the generation of high-producing cells. Using our CHOBRI/rcTA cell line, harbouring cumate gene switch components, we have selected in parallel two pools from the same transfection. One pool was induced at a high level during MSX selection (Cum1 pool), while the second pool was left non-induced (Cum0 pool), leaving only a basal hCD200Fc expression level due to the system’s leakiness. At the end of the selection recovery phase, pool productivity was compared in fed-batch culture carried out under the 38

Journal of Biotechnology 296 (2019) 32–41

A. Poulain, et al.

Fig. 4. Evaluation of μ-pools confluence and growth rate in 384-well plates. Cum0 and Cum1 pools were seeded at 5 cells/well in three 384-well plates and allowed to grow without MSX for ten days. Growth and production phase were decoupled in a biphasic culture process (panel A). The % confluence was calculated at the time of induction and at the end of the 5 days production phase. Percent confluence frequency distribution at induction time (panel B) and at the end of the production phase (panel C) for both Cum0 and Cum1 μ-pools is shown. Frequency distribution of μ-pools specific growth rate during the 5 days production phase was calculated based on the change in % confluence for Cum0 μ-pools and Cum1 μpools (panel D). The μ-pools growth efficiency was close to 100% for both pools since cell growth (confluence > 10%) at the end of the 5 days batch period was observed in 1146 (Cum0) and 1139 (Cum1) of the 1152 seeded wells. Only 6 Cum0 and 13 Cum1 μ-pools had confluence lower than 10%.

clearly indicates an inverse relationship between specific productivity and growth rate. Thus, during the selection process of the Cum1 pool, clones with high specific productivities likely became overgrown by low producers, thus reducing their frequency in the final pool. It is noteworthy that, by comparison with cells transfected with an empty pTT75 plasmid, hCD200Fc expression negatively affects the culture behavior: in the first few days following transfection and selection, when the plasmid copy number per cell is expected to be high, activation of r-protein expression leads to a reduction in cell proliferation and viability. In some studies, a correlation has been observed between high level of r-protein expression and ER-stress markers during transient expression (Johari et al., 2015; Ku et al., 2010). In response to ER stress, the UPR pathway can be activated in order to restore cellular homeostasis through various mechanisms such as attenuation of protein synthesis, expression of key components of the protein folding machinery or through cell cycle arrest in G1 (Johari et al., 2015; Ku et al., 2010; Zhou et al., 2018). When ER stress is prolonged, and homeostasis is not restored, the UPR can trigger apoptosis (Jager et al., 2012; Malhotra and Kaufman, 2007; Moore and Hollien, 2012; Zhou et al., 2018). Thus, the impact of hCD200Fc expression on the level of the endoplasmic reticulum (ER)-resident glucose regulated proteins GRP78/Bip, known to be an ER stress marker (Hendershot, 2004; Ku et al., 2010; Le Fourn et al., 2014), was investigated during the early days of the selection process. To discriminate between the stresses caused by r-protein overexpression from those caused by transfection and selection, Bip expression was also examined in cells transfected

the cell as it did not affect viability, its high-level expression seems to impose a metabolic constraint resulting in reduced cell growth. The fact that the Cum1 pool growth rate was not affected by hCD200Fc overexpression suggests that this pool is mainly composed of cells expressing lower levels of r-protein, ensuing reduced metabolic burden. To verify that the Cum0 pool is enriched in cells with higher specific productivity compared to the Cum1 pool, the frequency of high-producing cells within each pool was evaluated. Due to the low cloning efficiency of CHO cells observed following limiting dilution, and to reduce variability in cell recovery and growth generally observed after single cell cloning, μ-pools were isolated by seeding 5 cells per well in three 384-well plates. This allowed us to obtain 1146 (Cum0) and 1139 (Cum1) μ-pools showing > 10% confluence at Day 10, on a total of 1152 seeded wells, representing an outgrowth efficiency of > 98.9%. The % confluence distribution in each well at induction time (Day 10) was the same for Cum0 and Cum1 μ-pools, confirming that differences in cell growth are not a source of bias in our study. After a 5 days production phase following the addition of cumate, we observed that the proportion of Cum0 μ-pools producing high hCD200Fc titers was greater than Cum1 μ-pools. The Cum0 pool was also enriched in cells with higher specific productivity: more than 22% of Cum0 μ-pools had a relative specific productivity greater than 40 compared to only 3.3% in the Cum1 pools. Moreover, only one μ-pool with volumetric titer above 100 mg/L was observed in the Cum1 population, while 14 were present in the Cum0 pool. Importantly, the decreased specific growth rate observed in the induced Cum0 pool during the fed-batch culture 39

Journal of Biotechnology 296 (2019) 32–41

A. Poulain, et al.

compared to DUXB11 cells, resulted in a greater survival of high expressing clones (Hu et al., 2013). The authors hypothesised that the larger ER (and higher mitochondrial mass) in CHOK1 compared to DUXB11 cells may provide a better folding and secretion machinery and would allow for higher survival of high expressing clones during the selection process. Consequently, they proposed that cells not able to cope with ER stress would die during the selection process. Overall, our data strongly suggests that during MSX selection, cells constitutively expressing high levels of r-protein will experience increased levels of stress. High-expressers will be overgrown by fastergrowing/low-expressing cells while some high-expressing cells will not survive the combined burden of MSX selection and high-level protein expression. Thus, the use of an inducible system, that allows the selection process to be performed in the non-induced state, provides significant advantages in the isolation of high-producing clones by increasing their survival and frequency in a pool, thus facilitating clone screening and reducing cell line development time and costs. Author contributions Experiments were designed by AM, AP, BM, and YD. Experiments were performed by AP. AP wrote the manuscript, AM, BM and YD revised the manuscript. Declaration of interests A. Mullick and B Massie are co-inventors on the cumate-gene switch (patent US7745592B2). Acknowledgements The authors are grateful to Dr Matthew Stuible for critical reading of the manuscript and correcting English language. We thank Louis Bisson for performing protein A-HPLC analyses. This work was supported in part by the Natural Sciences and Engineering Research Council of Canada (NSERC). This is NRC publication #NRC-HHT_53407. References

Fig. 5. Pool selection in the non-induced mode increases the frequency of high producing cells. Ten days after seeding the Cum0 and Cum1 μ-pools in 3 × 384-well plates, cumate was added and hCD200Fc volumetric titer was quantified five days later in each well using a HTRF assay (panel A). Relative specific productivity was calculated using the % confluency at the end of the 5-days production phase (panel B). Box plot analysis are generated from the 1146 and 1139 μ-pools derived from the Cum0 and Cum1 pools, respectively. Boxes represent interquartile ranges (the median, the first and third quartile are indicated) and the cross indicate the average. The whiskers represent the maximum and minimum values excluding the outliers. Panel C illustrates the distribution (percentage) of μ-pools classified in five relative specific productivity categories.

Berting, A., Farcet, M.R., Kreil, T.R., 2010. Virus susceptibility of Chinese hamster ovary (CHO) cells and detection of viral contaminations by adventitious agent testing. Biotechnol. Bioeng. 106, 598–607. Broussau, S., Jabbour, N., Lachapelle, G., Durocher, Y., Tom, R., Transfiguracion, J., Gilbert, R., Massie, B., 2008. Inducible packaging cells for large-scale production of lentiviral vectors in serum-free suspension culture. Mol. Ther. 16, 500–507. Butler, M., Spearman, M., 2014. The choice of mammalian cell host and possibilities for glycosylation engineering. Curr. Opin. Biotechnol. 30, 107–112. Carpentier, E., Paris, S., Kamen, A.A., Durocher, Y., 2007. Limiting factors governing protein expression following polyethylenimine-mediated gene transfer in HEK293EBNA1 cells. J. Biotechnol. 128, 268–280. Durocher, Y., Butler, M., 2009. Expression systems for therapeutic glycoprotein production. Curr. Opin. Biotechnol. 20, 700–707. Durocher, Y., Perret, S., Kamen, A., 2002. High-level and high-throughput recombinant protein production by transient transfection of suspension-growing human 293EBNA1 cells. Nucleic Acids Res. 30, E9. Gaillet, B., Gilbert, R., Amziani, R., Guilbault, C., Gadoury, C., Caron, A.W., Mullick, A., Garnier, A., Massie, B., 2007. High-level recombinant protein production in CHO cells using an adenoviral vector and the cumate gene-switch. Biotechnol. Prog. 23, 200–209. Gaillet, B., Gilbert, R., Broussau, S., Pilotte, A., Malenfant, F., Mullick, A., Garnier, A., Massie, B., 2010. High-level recombinant protein production in CHO cells using lentiviral vectors and the cumate gene-switch. Biotechnol. Bioeng. 106, 203–215. Gilbert, R., Broussau, S., Massie, B., 2007. Protein production using lentiviral vectors. In: Dyson, M.R., Durocher, Y. (Eds.), Methods Express: Expression Systems. Scion Publishing Limited, Bloxham, Oxfordshire, UK, pp. 241–258. Hendershot, L.M., 2004. The ER function BiP is a master regulator of ER function. Mt. Sinai J. Med. 71, 289–297. Hetz, C., Papa, F.R., 2018. The unfolded protein response and cell fate control. Mol. Cell 69, 169–181. Hu, Z., Guo, D., Yip, S.S., Zhan, D., Misaghi, S., Joly, J.C., Snedecor, B.R., Shen, A.Y., 2013. Chinese hamster ovary K1 host cell enables stable cell line development for antibody molecules which are difficult to express in DUXB11-derived dihydrofolate reductase deficient host cell. Biotechnol. Prog. 29, 980–985. Jager, R., Bertrand, M.J., Gorman, A.M., Vandenabeele, P., Samali, A., 2012. The

with empty pTT75 plasmid. A small but significant increase in Bip expression levels was observed during the selection process when hCD200Fc expression was induced with cumate while cumate had no effect in empty pTT75-transfected cells. The fact that Bip expression was induced at Day 6 post-MSX addition in empty plasmid-transfected cells suggests that glutamine deprivation and MSX (a potent GS inhibitor) addition could induce an ER-stress as described previously (Shanware et al., 2014) Flow cytometry analyses following annexinV-FITC and propidium iodide (PI) labelling suggest that a higher proportion of cells overexpressing hCD200Fc died during the MSX selection process compared to non-induced cells. However, because the difference was only observed with annexinV-positive/PI-positive cells, corresponding to cells in late apoptosis or necrosis, we cannot conclude on the mechanism(s) leading to their death. These observations are consistent with a previous study where the use of CHOK1 cells, harbouring a larger ER size 40

Journal of Biotechnology 296 (2019) 32–41

A. Poulain, et al.

response. Semin. Cell Dev. Biol. 18, 716–731. Misaghi, S., Chang, J., Snedecor, B., 2014. It’s time to regulate: coping with productinduced nongenetic clonal instability in CHO cell lines via regulated protein expression. Biotechnol. Prog. 30, 1432–1440. Moore, K.A., Hollien, J., 2012. The unfolded protein response in secretory cell function. Annu. Rev. Genet. 46, 165–183. Mullick, A., Xu, Y., Warren, R., Koutroumanis, M., Guilbault, C., Broussau, S., Malenfant, F., Bourget, L., Lamoureux, L., Lo, R., Caron, A.W., Pilotte, A., Massie, B., 2006. The cumate gene-switch: a system for regulated expression in mammalian cells. BMC Biotechnol. 6, 43. Pilbrough, W., Munro, T.P., Gray, P., 2009. Intraclonal protein expression heterogeneity in recombinant CHO cells. PLoS One 4, e8432. Poulain, A., Perret, S., Malenfant, F., Mullick, A., Massie, B., Durocher, Y., 2017. Rapid protein production from stable CHO cell pools using plasmid vector and the cumate gene-switch. J. Biotechnol. 255, 16–27. Shanware, N.P., Bray, K., Eng, C.H., Wang, F., Follettie, M., Myers, J., Fantin, V.R., Abraham, R.T., 2014. Glutamine deprivation stimulates mTOR-JNK-dependent chemokine secretion. Nat. Commun. 5, 4900. Sommeregger, W., Mayrhofer, P., Steinfellner, W., Reinhart, D., Henry, M., Clynes, M., Meleady, P., Kunert, R., 2016. Proteomic differences in recombinant CHO cells producing two similar antibody fragments. Biotechnol. Bioeng. 113, 1902–1912. Wurm, F.M., 2004. Production of recombinant protein therapeutics in cultivated mammalian cells. Nat. Biotechnol. 22, 1393–1398. Zhou, Y., Raju, R., Alves, C., Gilbert, A., 2018. Debottlenecking protein secretion and reducing protein aggregation in the cellular host. Curr. Opin. Biotechnol. 53, 151–157.

unfolded protein response at the crossroads of cellular life and death during endoplasmic reticulum stress. Biol. Cell 104, 259–270. Johari, Y.B., Estes, S.D., Alves, C.S., Sinacore, M.S., James, D.C., 2015. Integrated cell and process engineering for improved transient production of a "difficult-to-express" fusion protein by CHO cells. Biotechnol. Bioeng. 112, 2527–2542. Kim, J.Y., Kim, Y.G., Lee, G.M., 2012. CHO cells in biotechnology for production of recombinant proteins: current state and further potential. Appl. Microbiol. Biotechnol. 93, 917–930. Kromenaker, S.J., Srienc, F., 1994. Stability of producer hybridoma cell lines after cell sorting: a case study. Biotechnol. Prog. 10, 299–307. Ku, S.C., Toh, P.C., Lee, Y.Y., Chusainow, J., Yap, M.G., Chao, S.H., 2010. Regulation of XBP-1 signaling during transient and stable recombinant protein production in CHO cells. Biotechnol. Prog. 26, 517–526. Lai, T., Yang, Y., Ng, S.K., 2013. Advances in Mammalian cell line development technologies for recombinant protein production. Pharmaceuticals (Basel) 6, 579–603. Lalonde, M.-E., Durocher, Y., 2017. Therapeutic glycoprotein production in mammalian cells. J. Biotechnol. 251, 128–140. Lam, C., Santell, L., Wilson, B., Yim, M., Louie, S., Tang, D., Shaw, D., Chan, P., Lazarus, R.A., Snedecor, B., Misaghi, S., 2017. Taming hyperactive hDNase I: Stable inducible expression of a hyperactive salt- and actin-resistant variant of human deoxyribonuclease I in CHO cells. Biotechnol. Prog. 33, 523–533. Le Fourn, V., Girod, P.A., Buceta, M., Regamey, A., Mermod, N., 2014. CHO cell engineering to prevent polypeptide aggregation and improve therapeutic protein secretion. Metab. Eng. 21, 91–102. Lee, G.M., Varma, A., Palsson, B.O., 1991. Application of population balance model to the loss of hybridoma antibody productivity. Biotechnol. Prog. 7, 72–75. Malhotra, J.D., Kaufman, R.J., 2007. The endoplasmic reticulum and the unfolded protein

41