One of the major challenges in development of producer cells for manufacturing recombinant proteins is the selection of high producer cells among a very large cell population, with different production levels. The selection of high producers is usually labor intensive and time consuming. The number of cells that can be screened manually is limited and so is the probability to isolate the clones of the required production level.
Traditional methods have been employed for screening for high producers. One of the common manual single cell cloning techniques is the “limiting dilution” procedure. Manual picking and selection of clones by this procedure is a tedious, labor intensive, and time consuming task. Even if performed by robotic systems, the number of clones that can be screened is limited ([1-3]). In addition, statistical analyses indicate that repeated cloning cycles are required to ensure clonality at an acceptable level ([4, 5]). Therefore, more efficient methods of cell line selection and cloning are required.
Flow cytometry () has been applied for selection of cells labelled with a fluorescent reporter. Some of the fluorescence activated cell sorting (FACS) machines are able to sort thousands of cells within a few seconds and select rare cells in the entire cell population (). Another advantage of flow cytometry is the potential for single cell cloning done by transferring single cells to separate wells of microtiter plates ([6, 8]).
However, cell sorting by flow cytometers requires a fluorescent signal. Because the protein product of interest is usually not fluorescent, a fluorogenic reporter gene product, whose expression is tightly linked to that of the gene of interest (GOI), is required.
The reporter gene product may be expressed in the cell cytoplasm or can be directed to the cell surface ([8, 9]). It can be either a naturally fluorescent protein such as green fluorescent protein (GFP) (), an internal protein specifically stained with a fluorescence marker that penetrates the cells such as fluorescent methotrexate (F-MTX) () or binds to a reporter molecule on the cell surface ([8, 9]).
An interesting case was reported for sorting antibody producing cells. It has been observed that the produced antibody molecule can be detected on the membrane of hybridoma cells ([12, 13]) as well as Chinese hamster ovary (CHO) cells (). Staining with fluorescently labelled anti-antibodies was used to detect the antibody on cell surface.
Other methods based on direct detection of the secreted protein of interest (POI) were also evaluated. These methods include gel microdrop technology ([15-17]) and matrix-based secretion assays ([18, 19]).
In addition, automated systems for cell selection also exist and include Laser enabled analysis and processing (LEAP) which destroys undesired cells ([20, 21]) and automated colony pickers such as ClonePix from Genetix and CellCelector™ from Aviso, which select directly the high producers according to their secretion levels. Beside the various pros and cons of each method, their capacity for the number of cells that can be screened is limited, and significantly smaller than that available by FACS.
On the basis of the above information we developed a new reporter gene with specific characteristics that can serve for cell sorting by FACS.
The reporter system described in this article fulfils the following characteristics: 1) specific staining with low background; 2) small size; 3) nontoxic; 4) Does not interfere with cell metabolism and growth rate; 5) Does not interfere with the ability of the cells to produce at high level a POI; and 6) uses the same secretion pathway of the reporter molecule as the POI.
Materials and Methods
CHO-S (GibcoBRL, Cat. # 11619, Grand Island, NY, USA) cells were adapted to grow in ProCHO5 medium (Lonza, BERD159Q, Verviers, Belgium).
Construction of Vectors
Vectors were constructed utilizing standard molecular biology techniques. The coding sequence for the BMP moiety fused to PAC via Insulin like growth factor 1 receptor (IGF1-R) TM domain (BMP-PAC) fragment was synthesized at GeneArt, Germany.
Vector MB-098 (pmCMV-LC-IRES-Qsy-mCMV-HC-IRES-BM/PAC-INS) and its membrane bound product are shown in Figures 1B and 1C. The vector was verified by restriction enzymes analysis and the BMP-PAC fragment was sequenced.
Preparation of Plasmid DNA
Plasmid deoxyribonucleic acid (DNA) was isolated using Hispeed plasmid Maxi Kit (Qiagen, 12663, Hilden, Germany).
The BMP-PAC fragment was sequenced at Hy-Labs, Israel, by the fully automated 16 Capillary ABI Prism 3100 Genetic Analyzer and analyzed by the Sci-Ed General software (CloneManager, version 7.01).
CHO-S cells were propagated in single cell suspension in ProCHO5 medium (Lonza, BERD159Q) in 50 ml filter tubes (TPP, 87050, Trasadingen, Switzerland), at 7°C, in 5% CO2, humidified and shaken 320 RPM. Two days before transfection, cells were seeded at a concentration of 0.2 × 106 cells/ml in 500 ml Erlenmeyer flasks (Corning, 431145, Lowell, MA, USA). On the day of transfection, 10 × 106 cells were seeded in 4 ml MEM (Sigma, M2279, Ayrshire, Scotland, UK) in T25 flasks (Nunc, 163371, Rokslide, Denmark). Linearized vector MB-098 (20 μg) (Figs. 1B and 1C) in a final volume of 100 μl MEM was mixed with 100 µl LipofectAmine (GibcoBRL, 18324-020, Carlsbad, CA, USA) and incubated for 45 minutes at RT. The mixture was added to the cells and incubated for 4 hours at 37°C, in 5% CO2 with shaking at 45 RPM. The cells were spun down and the medium was replaced with 20 ml fresh ProCHO5 (Lonza, BERD159Q) supplemented with Hypoxanthine 27.22 mg/L and Thymidine 7.76 mg/L (HTx2) (Biological Industries, 03-085-1C, Beit-Haemek, Israel) in 125 ml Erlenmeyer flask (Corning, 431143) and incubated at 37°C, in 5% CO2, with shaking at 125 RPM for 72 hours.
Cells were centrifuged and resuspended in 20 ml ProCHO5 medium supplemented with 20 µg/ml puromycin (Invitrogen, ant-pr-1, Toulouse, France), lacking L-glutamine and supplemented with GSEM (Sigma, liquid 50x, G9785 Ayrshire, Scotland, UK), 25 μM methionine sulphoximine (MSX) (Sigma, M5379), and 100 μg/ml dextrane sulfate (Sigma, D4911).
Cell Analysis by Flow Cytometry
The cells (2 × 106) expressing BMP-PAC were washed with phosphate-buffered saline (PBS) + 0.1% Pluronic acid F-68 (Sigma, P5556), incubated for 30 minutes, at 37°C and shaken at 80 RPM in PBS + 0.1% pluronic acid F-68 + R-Phycoerythrin-conjugated Streptavidin 2 μg/ml (BioLegend, 405203, San Diego, CA, USA) fluorescent streptavidin (F-SA) washed in PBS with 0.1% Pluronic acid F-68 and analyzed by BD FACS ARIA I (BD Biosciences, San Jose, CA, USA). The fluorescence excitation wavelength was 488 nm and the emission was measured with a filter of 578 nm using the blue Laser. For gating details see Gating, Supporting Information Figure 1S.
Cells were seeded (0.2 × 106 cell/ml) in 25 ml ProCHO5 into 50 ml tubes and incubated at 37°C on an orbital shaker (25 mm shaking radius) at 320 rpm. Twice a week, cell concentration and viability were measured (by Vi-Cell, Beckman Coulter, Nyon, Switzerland). Cells were passaged by centrifugation (100 g, 5 minutes, 4°C).
Cell Specific Productivity Determination in Animal Component Free Medium (ACFM)
Specific productivity, in picogram per cell per day (PCD), was determined by seeding cells in specified ACFM at a concentration of 0.5 × 106 cells/ml, in a 50 ml tube (TPP, 87050, Trasadingen Switzerland) and incubating on an orbital shaker (320 rpm) or in a T80 flask (Nunc, 153732, 45 rpm) at 37°C for 24 hours. Medium was sampled and the product concentration was determined by Enzyme-Linked Immunosorbent Assay (ELISA). Specific productivity was calculated by dividing the 24 hours titers by the average concentration of cells at seeding and after 24 hours.
Labeling of transfected cells, 40 × 106 cells/ml in T80 flasks, was done in PBS containing 0.1% pluronic acid F-68 (Sigma, P5556) and R-Phycoerythrin-conjugated streptavidin (F-SA) 2 μg/ml (BioLegend, 405203) for 1 hour, at 37°C with shaking at 80 rpm. After labeling, cells were washed twice in PBS + 0.1% Pluronic acid F-68 (Sigma, P5556), and resuspended at a final concentration of 10 × 106 cells/ml in PBS + 0.1% Pluronic acid F-68 (Sigma, P5556). The gated cells were sorted with the FACSAria flow cytometer (BD Biosciences) in “Single cell” precision mode as described below. The fluorescence excitation wavelength was 488 nm and the emission was measured with a filter of 578 nm using the blue laser. For gating details see Gating Supporting Information Figure 2S. Sorted cells were seeded in ProCHO5 medium and grown until cell viability was ≤90% and viable cell number was ≤70 × 106.
Cloning by Flow Cytometry Automated Cell Deposition Unit (ACDU)
Cloning of cells was done by the Automated Cell Deposition Unit (ACDU) device of the FACSAria cell sorter. A total of 2 × 106 cells were collected, washed twice in PBS + 0.1% pluronic acid (Sigma, P5556) and labeled in 0.5 ml of R-Phycoerythrin-conjugated Streptavidin (F-SA) 2 μg/ml (BioLegend, 405203). The cells were incubated in 24 well plates (Nunc, 143982) for 30 minutes at 37°C, by shaking at 80 rpm. Following labeling, cells were washed twice in PBS + 0.1% pluronic acid and resuspended in 4 ml of PBS + 0.1% pluronic acid (cell concentration ∼0.1–0.2 × 106 cell/ml). The highest fluorescent cells were cloned by the ACDU, as described in the Gating (Supporting Information Fig. 3S). The fluorescence excitation wavelength was 488 nm and the emission was measured with a filter of 578 nm using the blue laser. The highest fluorescent cells were sorted in “Single Cell” precision mode, into 96-well plates (Costar, 3595, Lowell, MA, USA) containing 180 μl/well of 80% Sigma C6366 medium and 20% ProCHO5 (BE12-766Q Lonza) ACFM mixture.
Formation of colonies in the microtiter plates' wells after cloning was monitored by the Cellavista instrument (automated image based platform technology, Innovatis, Roche) at day 0 and every 2–3 days, for detection of wells that contain only a single colony. Two weeks after cloning, supernatants from single colony containing wells were assayed by ELISA. Cells were picked from the wells with highest product titers and transferred to T25 flasks (Nunc, 163371) containing 4 ml of 50% Sigma C6366 Ayrshire, Scotland, UK and 50% ProCHO5 medium (Nunc) and incubated at 37°C without shaking. Cells were propagated in T25 flasks and then in T80 flasks (Nunc, 153732) with increasing volumes of ProCHO5 medium. Finally, cells were seeded in 25 ml fresh medium at 0.2 × 106 cells/ml for another growth cycle.
Cells (0.5 × 106 cells/ml) were seeded in 20 ml ProCHO5 medium in 50 ml filter tubes and cultured for 24 hours at 37°C at 320 rpm. The culture medium was centrifuged and filtrated (0.22 μm). The clarified supernatant was analyzed by ELISA and Western blot assays.
ELISA for Recombinant mAb
Microtiter plates (Nunc, MaxiSorp F96, Waltham, MA, USA) were coated with 100 μl per well of 2.0 μg/ml Goat anti-human Immunoglobulin (IgG) heavy chain+ light chain (Jackson, 109-005-088, West Grove, PA, USA) in coating buffer (Carbonate buffer: 100 mM, Na2CO3+NaHCO3, pH 9.6) and incubated overnight at 4°C in a humid box. The plates were washed four times with washing buffer PBS containing 0.05% of Tween 20 (Sigma, P-1379, St. Louis, MO, USA).
The plates were treated with 200 µl/well of blocking buffer containing 1% bovine serum albumin (BSA) (Bovostar, BSAS.01, East Keilor, Victoria, Australia) in 0.05% PBS-Tween20 for 1 hour at RT. The plates were washed four times with washing buffer and loaded with 100 μl aliquots of tested samples, standard curve (1.56–100 ng/ml) and check samples (with known concentration) in assay buffer [PBS 1x containing 1% Skim Milk (Fluka, 70166, Buchs, Switzerland)]. The plates were covered with a plate sealer and incubated for 1 hour at 37°C without shaking. Plates were washed again four times with washing buffer, and 100 μl of the second antibody, goat anti human IgG Fc (Fab)2 horse radish peroxidase (HRP) (Jackson Immunoresearch, 109-036-098) diluted 1:100,000 in assay buffer, were added to each well. The plates were incubated for 1 hour at 37°C. Plates were washed four times with washing buffer and 100 μl of substrate solution TMB (Savion Diagnostics, 1928, Ashdod, Israel) were added to each well and incubated 15–20 minutes at room temperature (without shaking). The reaction was stopped by adding 100 µl/well of a stop solution of 1N HCl (Merck, 1.00314.2500, Darmstadt, Germany). Absorbance was measured at A450 nm in an ELISA reader (TECAN, 16039400, Männedorf, Switzerland). Solutions of the recombinant mAb reference standard were prepared by serial dilutions in PBS, pH 6.0 to generate a standard curve in assay buffer ranging from 1.56 to 100 ng/ml (linear range between 1.56 and 50 ng/ml). A sample of crude harvest from the relevant pool was diluted with assay buffer to obtain ∼25 ng/ml and used as a control. The optical density data results were processed and results calculated by the Magellan software (WIZCON-PC). The dilution of samples, preparation of standard curve dilution series, and distribution of samples on the plate was performed by a robotic sample processor (TECAN, RSP 150/8).
SDS-PAGE/Western Blot Analysis
Clarified supernatant samples of recombinant mAb clones were diluted in Sodium Dodecyl Sulfate polyacrylamide gel electrophoresis (SDS PAGE) sample buffer. Samples of recombinant mAb clones (0.1 μg per lane by ELISA) were separated on SDS-PAGE 10 % Bis Tris gels (Invitrogen, NP0301BOX) under nonreducing conditions together with a 15 μl prestained molecular weight (MW) protein standard (Invitrogen, LC5925). Electrophoresis was performed at constant voltage (100 V) for ∼2 hours with a Mini-Cell gel electrophoresis system (Novex, EI9001, Carlsbad, CA). Proteins were transferred to a nitrocellulose membrane in a blotting module with transfer buffer at 35 volt for 1 hour. The membrane was rinsed and incubated with PBS-0.05% Tween 20 for 5 minutes, followed by incubation in 3% BSA (Bovostar, BSAS.01) in 0.05% PBS-Tween20 over night at 4°C. The membrane was incubated with the following antibodies diluted in 3% BSA in 0.05% PBS-Tween20, for 2 hours at room temperature with shaking: (a) Goat anti-Human IgG Fc HRP (Jackson, 109-036-098); (b) Goat anti-Human kappa light chain (Southern Biotech, 2060-01, Birmingham, AL, USA) followed by Donkey anti-goat HRP (Jackson, 705-015-147). The membrane was then washed for 5 minutes in PBS-0.05% Tween 20 three times. Bands were visualized by incubation in Enhanced Chemi Luminiscence (ECL) reagent (Pierce RPN 32106, Rockford, IL, USA) for 1 minute followed by exposure to film in an X-Ray Film (Fugi, F1824, Tokyo, Japan) Processor developer (Mini-Medical, AFP, Elmsford, NY, USA).
The platform (Fig. 1A) was designed in order to stringently select for cells producing high levels of the GOI. BMP was fused with the PAC resistance gene, thus generating a dual reporter and bioselection function. BMP-PAC enables bioselection with puromycin, followed by flow cytometry FCM analysis and FACS sorting according to the BMP levels. In a model system described in this article, a plasmid was constructed for the expression of a model recombinant mAb, composed of two expression cassettes one for the heavy chain and one for the light chain (Figs 1B and 1C). The heavy chain expression was driven by the powerful murine CMV IE2 (mCMV IE2) promoter and the BMP-PAC gene located downstream to an Encephalomyocarditis virus (EMCV internal ribosome entry site [IRES]). The light chain expression was driven by the murine CMV IE1 (mCMV IE1) promoter and a glutamine synthetase (QSy) selection marker located downstream to another EMCV IRES. This architecture dictates transcription of the heavy chain and BMP-PAC genes on a single mRNA and the transcription of the light chain and the QSy on another mRNA ().
The reporter-selection molecule contains the BMP preceded by a signal peptide derived from murine CD59a (). The BMP was fused at its C-terminus to a carrier composed of 60 amino acids, with two () potential N-glycosylation sites. This synthetic peptide was designed in order to be hydrophilic and long enough to protrude from the membrane and to expose the BMP to the extra cellular surface. The carrier was linked to a transmembrane (TM) domain from mouse IGF-I receptor. The TM domain was linked at its C-side to the PAC selection resistance gene via a short linker (Fig. 1B). Prediction of TM domain sequence was done by TMHMM server v.2.0.
The BMP is composed of the amino acid sequence: CHPQGPPC ([24, 25]).
PAC in this system is attached to the cytoplasmic side of cell membrane.
Generation and Evaluation of Stable Pools
CHO-S cells, transfected with the BMP-PAC containing vector MB-098, were subjected to selection with 20 µg/ml Puromycin and 25 μM MSX. Under these conditions, growth of the puromycin resistant stable cells indicates that PAC in the BMP-PAC chimera is active in the membrane bound form. Stable pools that express the recombinant mAb were selected and analyzed by FCM.
The stable producer cells were further sorted by FCM (Fig. 2). Analyses show that BMP was detected on the cell surface of the transfected cell population. Two subpopulations that express high and low BMP levels were seen in the pool before sorting. This result implies that selection by puromycin alone was not sufficient to isolate only the high producer cells. The analysis profile demonstrates that low BMP expressers can grow and propagate under the puromycin selection pressure conditions. This result confirms the necessity of BMP FACS sorting for more efficient selection of high producer cells.
Consecutive sorts performed by selection of the 4% highest BMP expresser cells at each sort resulted in elevation of the antibody specific productivity, which correlated with the BMP levels. The specific productivity of the antibody increased gradually with the sorting cycles from 4.0 PCD (pictogram per cell per day) before sorting, up to 10.8 PCD after the third sort. The specific productivity level correlated with the BMP level on the cell surface, as indicated by the fluorescent streptavidin, expressed in MFI units (mean fluorescence intensity) in Figure 2. The MFI before sorting cycles was 3273 and after the third sort was 26309.
The fact that no staining could be detected with CHO host nontransfected cells indicates the low background and high specificity of the method.
Cloning of Cells from Sorted Pools by the FACS ACDU Device
Following three consecutive sorting cycles, the top ∼3.2% highest fluorescent recombinant mAb producing cells, were cloned into 96-well plates. Colonies were monitored by the Cellavista instrument at seeding day and then every 2–3 days. Wells with more than one colony were discarded. From cloning onward the culture media did not contain puromycin.
Supernatants from 358 wells were sampled and assayed for productivity (titer) by ELISA (data not shown). Fifty one clones with the highest titers were transferred first to T25 flasks and then to T80 flasks for evaluation of specific productivity. The range of productivity was 2-26 PCD from which 34 clones had productivity above 10 PCD (Fig. 3).
Analysis of Protein Product from Clones
Protein product identification in crude cell culture medium by SDS PAGE Western blot
Selected clones were cultured as a single cell suspension. Product isolated from the crude cell culture media was analyzed by SDS-PAGE/Western blotting. Detection of the products was done with antibodies to human Fc and to the kappa light chain subunits (Figs. 3B and 3C, respectively).
The results show that the intact recombinant mAb with the apparent molecular weight of ∼150 kDa was identified by both detection antibodies. In addition, free light chain (LC) secreted by all clones (monomer and dimer according to apparent MW) was observed on the Western blots stained with the anti-LC (Fig. 5B).
Molecular weight of the intact product in the crude cell culture medium from all clones conforms to that of the purified product from the recombinant mAb reference sample. No significant difference in the apparent MW of the products of the different clones was found. Multiple bands may represent glycosylation microheterogeneity.
The Novel BMP-PAC chimera, composed of a membrane-bound reporter molecule fused to the PAC bio-selection gene was found to be a very efficient tool for high throughput sorting of high producer cells by FACS.
This bifunctional selection/reporter molecule enables to bioselect for stable transfected cells in the first step, and then to sort cells expressing the highest levels of the reporter gene and by that also to select the cells producing the highest level of POI.
The results demonstrate that the BMP reporter part of the BMP-PAC chimera could be detected and relatively quantified on the cell surface by FCM with high sensitivity and specificity, and with negligible background.
The PAC part of the chimera was found to be active as a cytoplasmic membrane bound protein, and therefore, endows the cells with the ability to propagate in the presence of puromycin.
The FCM BMP analysis showed low and high BMP expressing cell populations after selection with puromycin, implying that selection by puromycin is not sufficiently efficient by itself for isolation of high producers. This profile demonstrates that low BMP expressers can grow and propagate under puromycin selection pressure conditions. This phenomenon confirms the necessity of BMP FACS sorting for additional more efficient selection of high producer cells.
Consecutive sorting cycles of cells producing the highest reporter gene product levels resulted in correlated increase in the POI production. The F-streptavidin used in these experiments was confirmed to be of nonanimal derived origin, which enables the use this reagent in a clone development process, conforming to the guidelines of the health authorities in this respect.
Productivity increased significantly along the sorting cycles (Fig. 2). These results indicate that the expression of the reporter is tightly linked to that of the POI. Isolation of the highest producers was obtained by high throughput (HTS) screening of the isolated clones (Fig. 3).
In conclusion, the BMP-PAC chimera can be used as an efficient reporter gene for selection of high producer cells because of the following attributes: 1) it can be specifically stained with a negligible background; 2) it is relatively small in size; 3) the reporter is not toxic; 4) its expression does not reduce the ability of the cell to produce other recombinant proteins, as indicated by the positive correlation between the BMP-PAC chimera expression level and the POI productivity; 5) the reporter, as a membrane protein, has the same secretion pathway as that of the protein of interest, which could contribute to the positive correlation between the expressions rate of both; 6) no negative effect on the growth rate, metabolism, and viability of the cells was detected; 7) detection with the BMP-PAC is cost-effective because of the low, relatively expensive, F-streptavidin concentration required, 8) this reporter and method therefore enables fast high throughput sorting of high producer cells.