- Email: [email protected]
A simple method for enriching populations of transfected CHO cells for cells of higher specific productivity
Journal of Immunological Methods 277 (2003) 141 – 155 www.elsevier.com/locate/jim
Recombinant Technology
A simple method for enriching populations of transfected CHO cells for cells of higher specific productivity S.C.G. Brezinsky, G.G. Chiang, A. Szilvasi, S. Mohan, R.I. Shapiro, A. MacLean, W. Sisk, G. Thill * Biogen Inc., 14 Cambridge Center, Cambridge, MA 02142, USA Received 9 July 2002; received in revised form 18 October 2002; accepted 15 January 2003
Abstract To establish a simple and rapid method for the screening of stable recombinant Chinese hamster ovary (CHO) cell lines, we have developed a cell surface labeling technique using fluorescently tagged antibodies that bind to secreted target proteins at low temperature. Using fluorescence intensity as the sole criterion for selection of cells, we are able to enrich populations of highly productive cells using preparative flow cytometry sorting. Reiterative sorting based on selection of cells having the highest fluorescence intensity of cell surface labeled protein results in dramatic increases in specific cellular productivity. Using lymphotoxin-beta receptor IgG fusion protein as a model system, we have demonstrated a greater than 20-fold increase in specific productivity (0.49 – 11.5 pg cell 1 day 1) (pcd) without the use of methotrexate (MTX)-mediated selection or amplification. In addition, the flow cytometry used to enrich for and clone high producer cell lines has reduced development time by more than 50% and the number of screening assays by more than 10-fold. When a transfected population of CHO cells expressing a humanized version of the murine monoclonal antibody (mAb) AQC2 directed against human alpha 1 beta 1 integrin was subjected to the same treatment, a 25-fold improvement in specific productivity (0.3 – 8.0 pcd) was observed. Furthermore, similar application of this technique to MTX-amplified clones resulted in up to 120-fold overall improvement in specific productivity (up to 42 pcd). Greater than 20 examples are also presented to demonstrate the robustness and performance of this technique. D 2003 Elsevier Science B.V. All rights reserved. Keywords: Recombinant protein production; High-speed cell sorting; Cell surface labeling; Rapid cell screening; Productivity enrichment; Flow cytometry
1. Introduction
Abbreviations: CHO, Chinese hamster ovary; LTbetaR, lymphotoxin beta receptor; MTX, methotrexate; GFP, green fluorescent protein; DHFR, dihydrofolate reductase; RPE, R-phycoerythrin; PI, propidium iodide; SPR, specific productivity rate. * Corresponding author. Tel.: +1-617-679-3275; fax: +1-617679-3148. E-mail address: [email protected] (G. Thill).
Flow cytometry has provided a fast, simple method for identifying and isolating rare cell phenotypes from complex populations. However, the normal release of a secreted protein from the plasma membrane has created a challenge for flow cytometry-based cell line development. Secreted proteins are synthesized on ribosomes bound to the endoplasmic reticulum (ER). Inside the ER, proteins achieve their native conformation via
0022-1759/03/$ - see front matter D 2003 Elsevier Science B.V. All rights reserved. doi:10.1016/S0022-1759(03)00108-X
142
S.C.G. Brezinsky et al. / Journal of Immunological Methods 277 (2003) 141–155
mechanisms involving disulfide bond formation and post-translational modifications such as glycosylation. The attainment of the correct conformation is the ratelimiting step in protein secretion. Subsequently, the protein is transferred to the Golgi apparatus and then to the plasma membrane where the secreted protein is released from secretory vesicles through the plasma membrane into the surrounding milieu. Some of the recent methods attempting to compensate for the resulting secretion-dependent loss of product-specific staining include immobilization of the secreted product near each cell by either encapsulation (Powell and Weaver, 1990; Gray et al., 1995) or a cellular affinity matrix (Holmes and Al-Rubeai, 1999). While these methods have greatly improved the ability to identify and isolate high producers, the immobilization protocols are often difficult, time consuming, and limited in the number of cells that can be examined at one time. Another promising technique uses co-selection of a nonsecreted surrogate marker like green fluorescent protein (GFP) (Meng et al., 2000). As the association of the protein with the cell surface is transient and downstream of the rate-limiting step in the secretory process, we hypothesized that the protein on the cell surface should be detectable and might correlate with the amount of protein being secreted from the cell. If so, then we should be able to label such cells with a fluorescent reagent that binds to the secreted product and sort the higher productivity cells using flow cytometry. Using this hypothesis, we have developed a very simple, fast method of lowtemperature direct staining of proteins on the cell surface followed by sorting, without the use of special media or long incubation times. High-speed cell sorters can sort hundreds of millions of cells with exceptional accuracy, greatly enriching high producer populations, and can deposit one cell per well into 96-well plates to facilitate cloning. Three rounds of reiterative sorting followed by single cell seeding has resulted in clones with specific productivity 20 times higher than the unsorted transfected population. A second cell line development project that involved selection of clones followed by methotrexate (MTX) amplification resulted in >100-fold enrichment in specific productivity up to a specific productivity of 42 pcd. Additionally, we present timelines, screening effort, and final titers for 23 recombinant cell lines to demonstrate the overall robustness of the technique.
2. Materials and methods 2.1. Vectors All plasmids utilize the CMV intermediate early promoter. The promoter extends from a restriction site approximately 500 bp upstream of the TATA box to a polylinker near the initiation codon of the native CMV intermediate early structural gene. The promoter region includes splice donor and acceptor sites in the 5Vuntranslated region. The polyadenylation site is derived from a human growth hormone variant sequence. Plasmid pXLTBR.9 contains the lymphotoxin-beta receptor gene fused to human IgG domains, CH2 and CH3 (LTbetaR-Ig) (Browning et al., 1995), driven by the CMV promoter together with a wild-type dhfr gene expression cassette for selection. Plasmids pAND162 and pAND160 are light and heavy chain vectors, respectively, used for the expression of AQC2, a humanized monoclonal antibody (mAb). Plasmid pAND162 contains a neomycin resistance selectable marker, while pAND160 contains the wild-type dhfr gene. 2.2. Cells, cell culture, and transfections Dihydrofolate reductase (DHFR) deficient DG44 Chinese hamster ovary (CHO) host cells (Urlaub and Chasin, 1980) were maintained as suspension cultures in serum-free medium containing nucleosides. Transfections were carried out by electroporation. Transfected AQC2 mAb cell lines were grown in alpha minus MEM supplemented with 10% dialyzed fetal bovine serum (FBS) (Hyclone Laboratories, Logan, UT) and 2 mM glutamine (Life Technologies, Grand Island, NY). The LTbetaR-Ig cell lines were grown in HYQPF-CHO, a commercial serum-free medium from Hyclone Laboratories, or serum-free alpha plus MEM medium (alpha plus MEMSF), an enriched alpha plus MEM without FBS. Following electroporation, the cells were cultured in six-well tissue culture dishes (Becton Dickinson Labware, Franklin Lakes, NJ). For the LTbetaR-Ig fusion protein cell line, DG44 host cells were transfected with pXLTBR.9 and cultured in alpha plus MEMSF. Twenty-four hours post-transfection, the growth medium was changed to HYQPF-CHO. Six to eight days post-transfection, the wells were pooled and split into three T-75 tissue culture flasks (Corning, Corning,
S.C.G. Brezinsky et al. / Journal of Immunological Methods 277 (2003) 141–155
NY). Each flask was then treated as an individual population and scanned by flow cytometry. Populations with flow cytometry profiles showing high fluorescent intensity subpopulations were subsequently subjected to high-speed cell sorting into growth media supplemented with antibiotic and antimycotic drugs (Life Technologies). For the AQC2 mAb cell line development, DG44 host cells were co-transfected with plasmids pAND162 and pAND160. Three days post-transfection, 400 Ag/ml G418 (Genetecin) (Life Technologies) was added to the medium. Once cells had reached f 80% confluence, the wells were pooled, fed with fresh media, and sorted within 24 h. As a rule of thumb, both serum-containing media and serum-free media have been used successfully from electroporation to clone selection. In most cases where serum is used, cells are adapted to growth in suspension culture in serum-free media after clone selection. 2.3. Surface fluorescent antibody staining of recombinant CHO for secreted product Cells ready for sorting were harvested by Accutase treatment (Innovative Cell Technologies, La Jolla, CA) and then maintained at 0– 4 jC for all subsequent handling. Accutase is a collagenase from a non-animal source that is preferred for its potential advantages from a regulatory standpoint. Trypsin is a suitable substitute from a technical standpoint. The cells were passed through a 70-Am nylon mesh (Becton Dickinson Labware), washed twice with cold phosphatebuffered saline (PBS) (Life Technologies), and then counted and assessed for viability. The cells were pelleted again by centrifugation for 5 min at 1000 RPM at 4 jC, and resuspended in cold DMEM/2% BSA containing fluorescently labeled antibody. In general, 5 106 to 5 107 cells are labeled with 50 Ag of R-phycoerythrin (RPE) conjugated goat F(abV)2 anti-human IgG (Jackson Immunoresearch Laboratories, West Grove, PA) in Dulbecco’s modified eagle medium (DMEM) (Life Technologies), supplemented with 2% bovine serum albumin (BSA) (Sigma, St. Louis, MO). RPE conjugates are preferred to FITC conjugates since their higher quantum yield of fluorescence lends itself to better signal-to-noise ratio. This is especially true during the initial sort where the percentage of positive populations is at its lowest
143
relative to subsequent sorts. After a 15-min incubation on ice, the cells were washed twice with cold PBS and resuspended in PBS plus 2 Ag/ml propidium iodide (PI) (Molecular Probes, Eugene, OR). Approximately 5 105 cells were removed for pre-analysis by FACSCalibur cytometer (see below) before sorting. DG44 host cells were used as a negative control. For photomicrography, cells were stained with Alexafluor 647 (Molecular Probes) following the same staining protocol described above. Alexafluor 647 is preferred in this application because of its resistance to photobleaching relative to other fluorescent conjugates. After staining, the cells were resuspended in PBS without PI. 2.4. Flow cytometry and fluorescent-activated cell sorting Analytical flow cytometry scans were performed on a FACSCalibur flow cytometer (Becton Dickinson, San Jose, CA) equipped with Cellquest v3.0 software and an air-cooled argon laser emitting at 488 nm. The PE emission was detected on Fl-2 using a 585/42-nmband pass filter, and PI emission on Fl-3, using a 670nm-long pass filter. Dead cells were excluded in an FSC vs. Fl-3 dot plot, and PE-labeled signal gating was done on a live cell-gated Fl-2 histogram. High-speed cell sorts were performed on a Moflo flow cytometer (Cytomation, Fort Collins, CO), equipped with Summit 3.0 software and an argon laser emitting at 488 nm for fluorescence excitation. The PE emission was detected on Fl-2, using a 580/30-nm-band pass filter, and the PI emission was detected on Fl-3 using a 670/30-band pass filter. Compensation of PE/PI emission spectrum overlap was accomplished using Cytomation’s Digital Signal Processing (DSP) board in Summit. Dead cells were excluded in an FSC vs. Fl-3 dot plot, and doublets were excluded in an FSC width vs. area dot plot. PElabeled signal gating was done on a live cell-gated Fl-2 histogram. The sorting gate was the combination of the live cell gate, the doublet discrimination gate, and the histogram gate on Fl-2. 2.5. Confocal microscopy Fluorescent and differential interference contrast (DIC) photomicrographs were acquired on a Leica TCS SP confocal microscope equipped with a red
144
S.C.G. Brezinsky et al. / Journal of Immunological Methods 277 (2003) 141–155
laser diode and Leica confocal software V2.00 build 0368 (Leica Microsystems, Heidelberg, Germany). Photomicrographs were taken of cells observed through a 40 -oil immersion objective. 2.6. Enzyme-linked immunosorbent assay (ELISA) LTbetaR-Ig titers were determined from media samples by ELISA. Assay plates were coated with an LtbetaR specific antibody (Biogen). Bound LTbetaR-Ig was detected with anti-human IgG horseradish peroxidase (HRP) conjugate (Jackson Immunoresearch Laboratories). For the quantitation of AQC2
mAb by ELISA, assay plates were coated with an AQC2-specific antigen fusion protein. Bound AQC2 mAb was detected with donkey anti-human IgG (H + L) horseradish peroxidase conjugate (Jackson Immunoresearch Laboratories). For both assays, the concentration was determined by backfit from a curve using four-parameter fit with SoftMax (Molecular Devices, Sunnyvale, CA). 2.7. Western blot analysis Protein samples and standards (1 and 5 Ag/ml) in a 10-Al volume were mixed with an equal volume of
Fig. 1. Confocal photomicrography showing product-specific cell surface staining of an AQC2-secreting CHO cell line. Both the AQC2 cell line (upper panels) and the untransfected CHO host DG44 (lower panels) were stained with Alexafluor 647-conjugated anti-human antibody. Fluorescent staining is shown on the left-hand side while differential interference contrast is shown on the right-hand side. The cells were kept on ice until microscopic analysis.
S.C.G. Brezinsky et al. / Journal of Immunological Methods 277 (2003) 141–155
2 SDS-PAGE sample buffer and boiled for 3 min. All samples were loaded on a 4 –20% SDS-PAGE gel (Biorad, Hercules, CA) and run at 40 mA (constant current) until the bromphenol blue reached the bottom
145
of the gel. Proteins were transferred to polyvinylidene fluoride (PVDF) membrane (Biorad) in 10 mM CAPS, 10% methanol, pH 11 overnight at 25 V at 4 jC. The membrane was blocked for 1 h in 5% BSA
Fig. 2. (A) Histogram of fluorescent-activated cell sort, negative control. The Fl-2 histogram was derived from the combination of the live cell gate based on PI exclusion (FSC vs. Fl-3), and the doublet discrimination gate (pulse width vs. FSC), to exclude doublets. (B) First LTbetaR-Ig fluorescent-activated cell sort, pre-sort. Histogram of red fluorescence on Fl-2. A sort gate R2 (not shown) was set to collect the brightest 5% of R-PE positive cells for all three reiterative sorts.
146
S.C.G. Brezinsky et al. / Journal of Immunological Methods 277 (2003) 141–155
in Tris-buffered saline with 0.05% Tween-20 (TBST). The primary antibody was an alkaline phosphatase conjugate of a goat F(abV)2 anti-mouse IgG (Jackson Immunoresearch Laboratories) used at 1/10 000 dilution in 2% BSA/TBST. The incubation was conducted for 1 h at room temperature with agitation. The filter was then washed three times at room temperature for 10 min in 0.5% BSA/TBST with agitation. The alkaline phosphatase reaction was conducted in 100 mM Tris, pH 10, 100 mM NaCl, 1 mM MgCl2, with a 1/200 dilution of 25 mM CDP-Star (Roche Molecular, Indianapolis, IN) for 5 min prior to exposure to film for 1 –4 min. 2.8. Determination of specific productivity rate (SPR) For each population or clone, 1 105 cells were seeded per well of a six-well tissue culture plate (Corning) in 2-ml growth media. Assays were performed in triplicate. The cells were allowed to grow for 3 days, conditioned media harvested for analysis, and the cells were removed by Accutase and counted. Specific LTbetaR-Ig and AQC2 antibody titers were quantitatively determined from media samples by ELISA. The SPR measured in picograms of specific protein per cell per day (pg cell 1 day 1: pcd) is a function of both growth rate and productivity, as represented by the following equations:
SPR ¼
ICA ¼
Total protein mass ¼ qP Integral cell area ðICAÞ
ðfinal cell number initial cell numberÞ days in culture loge ðfinal cell number=initial cell numberÞ
3. Results 3.1. Direct, product-specific staining of secreted recombinant proteins at the plasma membrane with fluorescently labeled antibodies Transfected cells secreting recombinant proteins can be directly stained with fluorescently labeled antibodies directed against the heterologous protein. An example of this is shown in Fig. 1. CHO cells secreting the humanized version of the antibody AQC2 were labeled with an Alexafluor 647 derivative of an anti-human antibody. The staining was then viewed using laser confocal microscopy. As seen in Fig. 1 (upper left quadrant), intense staining of the cell surface is demonstrated. The differential interference contrast (upper right) shows the specificity of the staining for the cell surface. Fig. 1 (lower panels) shows the results of the same staining regimen for the untransfected CHO host, DG44. No staining is seen in this case. These data confirm that the heterologous secreted protein can be directly stained on the cell surface without additional immobilization, and that this protein-specific staining can be used as the basis for flow cytometry sorting. 3.2. Generation of high-producing recombinant CHO cell lines in the absence of methotrexate LTbetaR-Ig: The CHO host DG44 was transfected by electroporation with the LTbetaR-Ig expression plasmid, pXLTBR.9. After selection and outgrowth of the transfectants for more than 14 days, the cells were pooled and labeled with an RPE-labeled F(abV)2 fragment of a goat anti-human IgG molecule at 4 jC. These cells, along with a stained untransfected host negative control, were subjected to analytical flow cytometry prior to a preparative sort. Both a live/dead
Fig. 3. (A) Analytical scan of a sample of LTbetaR-Ig cells collected after the first sort, showing that the sort resulted in a very pure population with an increase in mean fluorescence intensity and a corresponding increase in specific productivity. Approximately 2 – 3 105 cells were recovered. The SPR values were acquired after expansion of the cells in culture. (B) Analytical scan of a sample of collected LTbetaR-Ig cells after second sort, demonstrating progressive increases in both fluorescence intensity and specific productivity after reiterative sorting. Approximately 2 – 3 105 cells were recovered. The SPR values were acquired after expansion of the cells in culture. (C) Analytical scan of a sample of collected LTbetaR-Ig cells after third sort, showing no additional increase in either mean fluorescence or specific productivity over second sort. Approximately 2 – 3 105 cells were recovered. The SPR values were acquired after expansion of the cells in culture.
S.C.G. Brezinsky et al. / Journal of Immunological Methods 277 (2003) 141–155
147
148
S.C.G. Brezinsky et al. / Journal of Immunological Methods 277 (2003) 141–155
gate (forward scatter vs. Fl-3) and a gate to eliminate doublets (pulse width vs. FSC) were used to evaluate the signals from the negative control (Fig. 2A) and the sample. The sample (Fig. 2B) contained populations of cells from which the fluorescence intensity greatly exceeded that of the negative control. For preparative sorting, a gate was set which encompassed cells with the top 5% of the fluorescence intensity of the sample population and sorted cells subsequently collected. The cell number was then expanded by culture under selective conditions, and the process was repeated two more times. For all sorts of LTbetaR-Ig producing cell lines, an analytical scan was performed post-preparative sorting to insure the quality of the sort. The analytical scan and the experimentally determined mean specific productivity of the pool are displayed in Fig. 3A –C and Table 1. The unsorted transfected cell population had a specific productivity of 0.5 pcd. Sorting based on fluorescence intensity improved the specific productivity average of the pools by approximately a 10-fold improvement to 5.1 pcd (see Fig. 3C). The third sort as measured either by fluorescence intensity or ELISA assay and SPR calculation failed to demonstrate an increase in the specific productivity of the pool (Fig. 3B and C, Table 1). In fact, the mean fluorescence intensity (MFI) is lower relative to the second sort after the sorted cells from the third sort were subjected to analytical flow cytometry post-sort. Based on our experience, we ascribe no significance to this downward shift in MFI. It is related to variability in cytometer perform-
Table 1 Specific productivity rates (SPR) of unsorted and consecutively sorted CHO pools of LTbetaR-Ig demonstrate that specific productivity increases with reiterative sorting LTbetaR-Ig pool
Mean fluorescence intensity
Average SPR (pg cell 1 day 1) F S.D.
Unsorted First sort Second sort 2 Third sort 3
12.91 33.39 143.78 86.35
0.5 F 0.0 3.1 F 0.1 4.5 F 0.3 5.1 F 0.7
For the SPR assay, 1 105 cells were cultured in a six-well tissue culture dish in 2-ml serum-containing media. The cells and supernatant were harvested after 3 days in culture. The SPR assay was conducted in triplicate for each sample.
ance from run to run. This interpretation is supported by the SPR data that show a minimal increase in specific productivity. During the third sort, single cells were isolated directly from the cytometer into 96-well plates. The clonality of the seeded wells was verified by visual inspection. Several weeks later, clones were assayed for specific productivity and found to lie with a range of specific productivity between 3.6 and 11.5 pcd. The most productive clone represents a 23-fold enrichment in specific productivity without the need for MTX amplification. The number of clones assayed by ELISA during this enrichment was 50. The timeline was approximately 9 weeks from transfection to identification of the best clone. 3.3. Generation of methotrexate-amplified recombinant CHO cell lines AQC2 mAb: A second example of this technique is the expression of a humanized antibody to human alpha 1 beta 1 integrin. This approach underscores both the utility of this technique to enhance selection of high-producing clones and its usefulness in enriching and subcloning pools of MTX-amplified lines in order to meet stringent titer goals. The light and heavy chain were expressed from separate plasmids similar to that described for LTbetaR-Ig. After transfection and expansion under DHFR and G418 selection, the entire population, with a specific productivity of 0.3 pcd, was labeled with a fluorescent F(abV)2 fragment of goat anti-human IgG, and the top 2 –5% expressing cells as measured by fluorescence intensity were collected. After approximately 1 week of expansion, sorted cells were subjected to another sort. The cells were expanded again, then deposited at one cell per well into 96-well plates during the third sort, followed by visual verification of clonality. As in the case of LTbetaR-Ig, sorting resulted in a steady increase in the fluorescence intensity of the labeled cells as well as the measured specific productivity of both pools and clones (data not shown). Approximately 117 clones were expanded into 24-well plates and screened for antibody titer. The specific productivity of the highest expressing clones was determined in the SPR assay (Table 2). As the specific productivity of the best clones was not adequate for our needs, we performed a methotrexate
S.C.G. Brezinsky et al. / Journal of Immunological Methods 277 (2003) 141–155 Table 2 Specific productivity rates of AQC2 mAb CHO cells isolated from pools subjected to three reiterative FACS sorts before and after MTX amplification nM MTX Unamplified parent clone Amplified pool Amplified subclone Unamplified parent clone Amplified pool Amplified subclone
Amplified pool Amplified subclone
5A
SPR (pg cell 1 day1)
–
3.3
5AB 5AB-17 5AB-52 11B
250 250 250 –
16.6 41.0 32.3 8.0
11BB 11BB-46 11BB-67 11BB-68 11BB-83 11BA 11BA-1 11BA-30 11BA-41 11BA-47 11BA-50 11BA-118 11BA-100
250 250 250 250 250 100 100 100 100 100 100 100 100
13.5 25.4 27.3 19.9 26.9 17.9 18.4 19.9 18.2 26.6 26.1 28.1 32.5
No. of clones screened from amplified pool
52
121
126
Triplicate 3-day SPR assays were conducted in serum-containing media for each sample.
149
rescent-activated cell sorting and cloning, the best producers from 52 clones assayed exhibited qPs of 41.0 and 32.3 pcd. Similarly for clone 11B, cells amplified in 100 nM MTX had a mean specific productivity of 18 pcd and produced clones of up to 32.5 pcd out of approximately 126 clones screened (Table 2). The 250 nM MTX-amplified pool of clone 11B had a qP of 13.5 pcd and produced clones of up to 27 pcd in a similar size screen (Table 2). Fig. 4 shows a dramatic increase in the mean fluorescence intensity ( f 32-fold) of the top 100 nM MTX-amplified clone, 11BA-100, compared to the initial pool of transfectants, which correlates to the high-level product secretion ( f 110-fold increase in specific productivity). The increase in fluorescence intensity correlates to the dramatic increase in protein secretion. As was seen in the case of the LTbetaR-Ig fusion protein, fluorescence intensity is a very good surrogate marker for specific cellular productivity. This parameter can be used to guide cell line selection without handling individual clones during the selection process and without heavy reliance on ELISA assays. 3.4. Sorting improves per cell productivity
amplification on several clones. G418 was removed from the 10 highest expressing clones, prior to their being subjected to amplification in media containing either 100 or 250 nM MTX. Based on our experience, we favor methotrexate concentrations in the 25 –250 nM range as they are less likely than higher concentrations to slow the growth rate of the production cell line but can still generate specific productivity increases. Amplified pools were screened for antibody titer. Populations exhibiting a qP equal to or greater than 13.5 pcd were subjected to highspeed cell sorting and autocloning of the highest 2% expressing population into 96-well plates. Subcloning is required at this point because methotrexate amplification is a stochastic process generating heterogeneity in the treated population. Two of the top antibody-producing clones, clone 5A and 11B, had unamplified qPs of 3.3 and 8.0 pcd, respectively (Table 2). When clone 5A was amplified in the presence of 250 nM MTX, a pool of specific productivity of 16.6 pcd was generated. After fluo-
As a control experiment to demonstrate the utility of cell surface labeling and flow cytometry sorting in enriching populations of cells for high producers, we compared the following two samples. The first is a non-amplified transfection expressing a human cytokine receptor fused to human Fc (Table 4, line 3) sorted multiple times. After cloning, 94 out of 106 (89%) of clones sorted two or three times produced greater than 1 Ag/ml in an assay of 24-h old media supernatants. We found that 45% of the positive clones produced >5 Ag/ml and 9% produced >10 Ag/ml. Table 3 shows the distribution of ELISA titers in this experiment. When a collection of randomly selected clones from a similar transfection treated with 25 nM MTX to boost productivity was similarly assayed, only 8 out of 100 (8%) were positive by the same criterion. All clones produced between 1 and 5 Ag/ml. A second control experiment was performed as follows. A cell line was identified (Table 4, line 18) that had been shown to produce Western blot positive
150
S.C.G. Brezinsky et al. / Journal of Immunological Methods 277 (2003) 141–155
Fig. 4. Analytical scan of unsorted AQC2 mAb CHO pool (black) vs. amplified clone 11BA-100 (green) demonstrates an increase in both mean fluorescence intensity and specific productivity after MTX amplification, sorting, and cloning. The Fl-2 histograms were derived by analysis of the PE signal within the live cell gate.
clones (>1 Ag/ml by visual inspection in 8 out of 8 cases and >5 Ag/ml in 7 out of 8 cases). In order to demonstrate the magnitude of the enrichment produced by cell surface labeling and sorting, this line was resorted from a frozen vial of previously unsorted
Table 3 Enrichment of clones for titer using fluorescent labeling and flow cytometry Criterion (Ag/ml)
Random clone selection
Two to three enrichment sorts
EXAMPLE #1a <1 1–5 5 – 10 >10
92/100 8/100 0 0
12/106 54/106 32/106 8/106
EXAMPLE #2b <1 1–5 >5
24/26 2/26 0
0 8/8 7/8
a Samples were measured by ELISA after a 24-h incubation in fresh media. b Samples were measured after a 24-h incubation in fresh media in a Western blot format and assessed by visual inspection.
cells from the same transfection. Following labeling, cells were sorted into 96-well plates with the Fl-2 sort gate set at 100%, i.e., without enrichment. Following outgrowth, media conditioned for 24 h from nearly confluent plates was assayed by Western blot and compared to a 1 Ag/ml standard. We found that 2 out of 26 lines analyzed (8%) met the 1 Ag/ml criterion and barely so (data not shown). Upon prolonged exposure, the protein could be seen in about half of the samples, but was very much reduced in intensity relative to the 1 Ag/ml standard. In both cases, sorting improved both the percentage of positive clones and the titers produced by the positive clones relative to unsorted cells (Table 3). 3.5. Historical list of applications Table 4 lists a series of 23 cell lines developed using this method. A generic description of the protein is noted in the first column of the table. The Fc portion of all the receptor-Fc fusion proteins is either human IgG1 or murine IgG2b. Two examples, lines 5 and 16, which were not receptor-Fc fusion proteins, used murine monoclonal antibodies followed by an RPE-
S.C.G. Brezinsky et al. / Journal of Immunological Methods 277 (2003) 141–155 Table 4 Results obtained from multiple cell lines using flow cytometry to enrich for increased per cell productivity Product
No. of Final No. of Timeline sorts titer assays to clone mg/la ID (weeks)
(1) (2) (3) (4)
3 2 3 2
150 140 200 180
100 20 106 20
12 7 7 8
2
50
11
7
2
40
24
7
2
100
20
7
2
50
20
8
2
33
10
7
2
70
10
6.5
2 2 2 2 3 2
125 55 100 75 40 75
10 10 20 10 150 20
7 7.5 8 7 9 9
2 2 2
60 160 70
20 8 20
6 7 6.5
3 2
166 130
148 18
8 8
2 3
150 242
18 129
7 8
Immune system ReD/hFc Type III ReD/hFc Hormone ReD/hFc Secreted version of GPI-linked protein/FC (5) Murine adhesion molecule/Flag tag (6) Murine immune system ReD/hFc mutantA (7) Murine immune system ReD/hFc mutantB (8) Human immune system ReD/hFc mutantA (9) Human immune system ReD/hFc mutantB (10) Human immune system ReD/hFc mutantC (11) Human ReD/Fc (12) Mutant of ReD/Fc A (13) Mutant of ReD/Fc B (14) Mutant of ReD/Fc (15) Mutant of ReD/Fc (16) Secreted version of GPI-linked protein/no tag (17) Cytokine ReD/Fc (18) Murine ReD/MigG2b (19) Murine kidney surface molecule-hFC (20) Antibody (21) Murine immune system ReD/mFc (22) MReD-hFc (23) Antibody #2
Receptor extracellular domain (ReD). a Cells were adapted to growth in suspension in serum-free media. Titer was determined by A280 following protein A capture in most cases.
labeled anti-mouse conjugate for detection. Cell lines used for preliminary research experiments, e.g. in vitro experiments or small animal disease models, are subjected to two sorts and a small number of assays are performed. Western blots of media conditioned for 24 h are measured relative to standards (1 and 5 Ag/ml). This assay is sufficient to permit the choice of a line capable of generating enough protein for drug candidate evaluation studies. Cells destined for commercial purposes are sorted and assayed more extensively as
151
higher titer may be necessary. In most cases, 6 to 8 weeks is all that is needed to identify a number of clones capable of producing titers of 50– 200 mg/l. Methotrexate is not generally used in the construction of these lines as it requires additional time to amplify and select clones. The only difficult protein on the list is #15 which is the last in a series of mutants (11 –15) and the only one that did not easily yield good clones. This protein seemed to be adhering to the cell surface, i.e., was FACS positive but without secretion into the medium in almost all cases. However, when a positive clone was found, it was amplified and clones producing >150 mg/l were easily found. We have noted failures using this approach related to ‘‘protein engineering’’ problems, e.g. attempts to pentamerize antibodies (three examples). We also had two examples where the protein was toxic (a large amount of cell death accompanied both the initial transfection and subsequent enrichment steps) to cells and did not yield productive clones. In addition, the use of commercial vectors, e.g. pCMV Script (Stratagene, La Jolla, CA) is not advisable. This technique is very sensitive to the ability of the cell to deliver the desired protein to the cell surface at a high enough level to produce a signal. Transcriptionally weak vectors do not produce enough protein to do so. We strongly recommend that only vectors such as the one described herein be used for the purposes of flow cytometry-based cell line development. 3.6. Other applications We have investigated the applicability of this technique to other cell types that produce proteins of interest to the biotechnology field. We labeled both a hybridoma (murine monoclonal IgG1) against a human cell surface protein and its fusion partner with an RPE-labeled conjugate directed against heavy and light chains. Results are shown in Fig. 5A. The black and green curves show the fusion partner in the absence and presence of staining. Mean fluorescence intensities are comparable. The cultured hybridoma reacted strongly with the antimouse IgG conjugate (red curve) and had a mean fluorescence intensity almost 60-fold greater than background. It also appears to contain at least two populations of cells, one a low abundance high-
152
S.C.G. Brezinsky et al. / Journal of Immunological Methods 277 (2003) 141–155
Fig. 5. Histograms of red fluorescence intensity. There were 1 – 2 106 cells labeled with 2 Ag RPE-labeled goat anti-mouse IgG (H + L). (A) FL653 fusion partner unstained (black), FL653 fusion partner stained (green), A8 G3.5 monoclonal antibody (red). (B) NS/0 host unstained (black), NS/0 host stained (green), mouse anti-human CD154 antibody (red).
expressing group and a more abundant population with a much lower MFI. As a second example of alternative hosts, we examined the expression of a murine anti-human CD154 antibody from transfected NS/0 cells using the cell surface staining technique. The nontransfected host (Fig. 5B) had an MFI of 1.6, while the stained version of the host was about twofold higher. This indicates a background level of antibody or light chain secretion from or cell surface localization on the host. The transfected host expressing the antibody had an
MFI greater than 30-fold showing that it was clearly positive. The most surprising result though was the breadth peak of MFI in the histogram. We do not see such broad peaks for clonal lines in CHO cells.
4. Discussion The screening method described herein, for transfected cell lines producing heterologous proteins, relies on the fact that the secreted protein is transiently
S.C.G. Brezinsky et al. / Journal of Immunological Methods 277 (2003) 141–155
associated with the cell surface during protein secretion. As such, the cell can be labeled with a fluorescent reagent such as a protein-specific antibody. Fluorescently labeled cells are excellent substrates for flow cytometry. A study by Kromenaker and Srienc (1994) involving direct staining and sorting of hybridomas indicated that fluorescence of surfaceassociated IgG is stable for an extended period of time. These data correlate with our own findings that the fluorescence of directly IgG-stained CHO producer cells is stable for over 1 h at room temperature as measured by flow cytometry and laser confocal microscopy (unpublished observations). The effectiveness of this method is demonstrated by the fact that, in the two cases presented in detail in this paper, fluorescence intensity of cell surface proteins was used as the predominant criteria for selection of clones and resulted in the selection of clones with relatively high-specific productivity. A similar direct relationship has been found between surface IgG expression and antibody secretion rate in hybridomas, although this is not true for all cell lines (Sen et al., 1990; Cherlet et al., 1995; Marder et al., 1990; McKinney et al., 1991). The technique has several advantages from a handling standpoint. First, it is thorough since all cells present in the transfected or amplified populations are examined. Second, all cells up to the point of cloning are handled in batch. As very little handling is involved, it is possible for a single person to produce multiple cell lines. Immunoassay labor is also greatly reduced as initial screening is performed by the cytometer. Relative to techniques such as Gel Microdrop technology, this technique is simpler and more reliable (unpublished observations). In addition, at over 100 million cells sorted after a single staining event, we have not yet reached the limit on the number of cells capable of being processed with this protocol (unpublished observations). In addition, the continuing improvements in fluorescent-activated cell sorting technology have improved the accuracy of population enrichment and single cell seeding. The current methodology directly measures the presence of the secreted heterologous protein on the cell surface and is potentially superior to surrogate markers such as co-transfected green fluorescence protein (GFP). GFP is a good measure of transcription, but secretion of heterologous proteins is not
153
necessarily a linear process. Overexpression of protein can deplete oxidizing reagents involved in proper folding of proteins leading to misfolding and inefficient secretion (Parekh and Wittrup, 1997). Hence, increased transcription does not necessarily correlate with increased secretion. Our method selects for the cell line that possesses the proper cellular machinery and conditions (integration site or MTX concentration, etc.) that lead to maximal presence on the cell surface, which is a better predictor of secretion. The two described examples are representative of many of applications of this technique (see Table 4). In the first case, an unamplified line was generated in approximately 9 weeks from transfection to identification of the best clone. The timeline for MTX amplification about twice as long as sorting is repeated after several clones are subjected to multiple MTX concentrations. This fluorescent-activated cell sorting method is also a good analytical method for screening several clones under multiple MTX concentrations and choosing the best combination of both. As shown in Table 4, a large number of cell lines have been developed using this technique. These lines were developed in a relatively short period of time. Mean development time is 7.6 weeks for clone selection. Average assay number is 43 assays per line. Greater than 50% of the clones produced the protein at >100 mg/l. We have performed two control experiments to show that cell surface labeling followed by sorting enriches the number of positive clones for a given expression criteria. In both experiments, the increase in number of positives was >10-fold. In addition, among the positive clones generated by this method, the average titer was significantly higher (see Table 3). Preliminary experiments using this technique on hybridoma lines and transfected NS/0 cells indicate that the method can detect the presence of secreted antibody in these hosts as well as in CHO transfectants. This technique may be useful in hybridoma subcloning, isotype selection, or in monitoring the stability of expression during routine passage. The hybridoma line had two separate populations, the less abundant having the higher MFI. This may be a reflection of selection pressure against the high producer population mediated by the genetic instability of the myeloma line. A second phenomenon of note was the small increment of background staining in the NS/0 line. Myeloma lines are
154
S.C.G. Brezinsky et al. / Journal of Immunological Methods 277 (2003) 141–155
professional antibody producing cells which, coupled with their unstable genome, could lead to spontaneous antibody production. Such background signals would clearly interfere with selection of high-producing clones. Myeloma lines are qualitatively different from CHO cells being lymphoid cells as opposed to the CHO epithelial lineage. This may impact the nature of the glycocalyx (see discussion below on mechanism) and require a more thorough investigation of cell culture and fluorescent labeling conditions in order to optimize this technique, Nevertheless, at first approximation, this technique seems to be capable of detecting antibody secretion in both hosts. Although we have not thoroughly explored the mechanism by which the secreted protein is held onto the cell surface and easily detected, we believe that there are two leading hypotheses: (1) staining of the secreted protein on the plasma membrane as it is being released from secretory vesicles or (2) transient association of the secreted protein with the protein or proteoglycan components of the glycocalyx (Bennett,1963). We favor the glycocalyx hypothesis for several reasons. The first relates to stability of signal, which is greater than 1 h. During secretion, the protein is packaged into secretory vesicles and rapidly transported to the plasma membrane. Vesicles fuse on the cytoplasmic side of the membrane, and the contents of the secretory vesicle cross the membrane through an aqueous pore (Lippincott-Schwartz et al., 2000; Scales et al., 2000). We have investigated the temperature dependence of labeling to see if labeling the cells at 4 jC, due to perturbation of membrane fluidity, is a component of signal generation and strength. Comparable signals were obtained at 4, 23, and 37 jC (unpublished observations). Given the temperature independence of cell surface labeling coupled with the rapidity of secretory vesicle fusion to the plasma membrane, and in contrast to the longevity of the signal, we doubt that we are observing labeling of the protein on the plasma membrane. Second, there is a precedence for binding of secreted proteins to the glycocalyx. This phenomenon of cell surface attachment for secreted proteins has been observed for FGF-2 (Engling et al., 2002). FGF2 is delivered to the cell surface via a nonclassical secretion mechanism and is found attached to surface proteoglycan.
Third, we believe that the glycocalyx hypothesis explains one of the more puzzling aspect of the technology, false positives. Every cell line development project encounters some percentage of false positives. False positives are clones with high mean fluorescence intensity values but no protein secretion. Table 4, lines 11 –15, shows a series of mutants of an immune system receptor-Fc fusion protein which posed no problems in cell line development save for cell line #15. This molecule showed an unusually high percentage of false positives (>95%). When some of these nonsecretors were reinterrogated for fluorescence intensity, MFI was well above background. Our explanation was that the protein was produced at a low level and sticking to the cell surface producing a fluorescent signal but not being secreted. Only two clones secreting the protein were ever found out of f 150 screened. One of these clones was capable of being amplified by MTX (50 nM) treatment. It produced high levels of protein in 14 out of 14 clones surveyed. These observations are consistent with proteins being localized to the cell surface via interaction with the cell surface glycoproteins. Once a binding threshold was exceeded, secretion was facile. These observations are not easily explained with a plasma membrane/secretory vesicle model, especially when considering the saturation aspects of the phenomenon. Lastly, the stability of the sorted populations, as reflected in change in mean fluorescence intensity between sorts, is a good indicator of the physiological impact of expression of a given heterologous protein on the host. The physiological impact ranges from excellent stability to severe collapse of the mean fluorescence intensity. The latter case may reflect toxicity of the foreign protein to the host, differential growth rates between cells with different expression levels, or slow adaptation of high producers when cultured at low density. Knowledge of these phenomena can assist in planning screening strategies (number of sorts, number of clones to assay, choice of slow vs. fast growers).
Acknowledgements The authors acknowledge the support of Konrad Miatkowski during the development of this technology. We thank Stephan Miller for guidance and
S.C.G. Brezinsky et al. / Journal of Immunological Methods 277 (2003) 141–155
assistance in manuscript preparation. We are also grateful to John McCoy for critical comments on the content of the manuscript.
References Bennett, H.S., 1963. Morphological aspects of extracellular polysaccharides. J. Histochem. Cytochem. 11, 14 – 23. Borth, N., Zeyda, M., Katinger, H., 2000. Efficient subcloning of high-producing subclones during gene amplification of Chinese hamster ovary cells by flow cytometry and cell sorting. Biotechnol. Bioeng. 71, 266 – 273. Browning, J.L., Dougas, I., Ngam-Ek, A., Bourdon, P.R., Ehrenfels, B.N., Miatkowski, K., Zafari, M., Yampaglia, A.M., Lawton, P., Meier, W., Benjamin, C.P., Hession, C., 1995. Characterization of surface lymphotoxin forms. Use of monoclonal antibodies and soluble receptors. J. Immunol. 154, 33. Cherlet, M., Kromenaker, S.J., Sirenc, F., 1995. Surface IgG content of murine hybridomas: direct evidence for variation of antibody secretion rates during the cell cycle. Biotechnol. Bioeng. 47, 535 – 540. Engling, A., Backhaus, R., Stegmayer, C., Zehe, C., Selenmeyer, C., Kehlenbach, A., Schwappach, B., Wegehingel, S., Nickel, W., 2002. Biosynthetic FGF-2 is targeted to non-lipid rafts following translocation to extracellular surface of CHO cells. J. Cell Sci. 115, 3619 – 3631. Gray, F., Kenny, J.S., Dunne, J.F., 1995. Secretion capture and report web: use of affinity derivatized agarose microdroplets for selection of hybridoma cells. J. Immunol. Methods 182, 155 – 163. Holmes, P., Al-Rubeai, M., 1999. Improved cell line development by a high throughput affinity capture surface display technique to select for high secretors. J. Immunol. Methods 230, 141 – 147. Kaufman, R.J., Sharpe, P.A., 1982. Amplification and expression of sequences cotransfected with a modular dihydrofolate reductase complementary DNA gene. J. Mol. Biol. 159, 601 – 621. Kim, N.S., Byun, T.H., Lee, G.M., 2001. Key determinants in the occurrence of clonal variation in humanized antibody expression
155
of CHO cells during dihydrofolate reductase mediated gene amplification. Biotechnol. Prog. 17, 69 – 75. Kromenaker, S.J., Srienc, F., 1994. Stability of producer hybridoma cell lines after cell sorting: a case study. Biotechnol. Prog. 10, 299 – 307. Lippincott-Schwartz, J., Roberts, T.H., Hirschberg, K., 2000. Secretory protein trafficking and organelle dynamics in living cells. Annu. Rev. Cell Dev. Biol. 16, 557 – 589. Marder, P., Maciak, R.S., Fouts, R.L., Baker, R.S., Starling, J.L., 1990. Selective cloning of hybridoma cells for enhanced immunoglobulin production using flow cytometric cell sorting and automated laser nephelometry. Cytometry 11, 498 – 505. McKinney, K.L., Dilwith, R., Belfort, G., 1991. Manipulation of heterogeneous hybridoma cultures for overproduction of monoclonal antibodies. Biotechnol. Prog. 7, 445 – 454. McKinney, K.L., Dilwith, R., Belfort, G., 1995. Optimizing antibody production in batch hybridoma culture. J. Biotechnol. 40, 31 – 48. Meng, Y.G., Liang, J., Wong, W.L., Chisholm., V., 2000. Green fluorescent protein as a second selectable marker for selection of high producing clones from transfected CHO cells. Gene 25;242 (1 – 2), 201 – 207. Parekh, R.N., Wittrup, K.D., 1997. Expression level tuning for optimal heterologous protein secretion in Saccharomyces cerevisiae. Biotechnol. Prog. 13 (2), 117 – 122. Powell, K.T., Weaver, J.C., 1990. Gel microdroplets and flow cytometry: rapid determination of antibody secretion by individual cells within a cell population. Bio/Technology 8, 333 – 337. Scales, S.J., Bock, J.B., Scheller, R.H., 2000. The specifics of membrane fusion. Nature 407, 144 – 146. Sen, S., Hu, W.-S., Srienc, F., 1990. Flow cytometric study of hybridoma cell culture: correlation between cell surface fluorescence and IgG production rate. Enzyme Microb. Technol. 12, 571 – 577. Urlaub, G., Chasin, L., 1980. Isolation of Chinese hamster ovary cell mutants deficient in dihydrofolate reductase activity. Proc. Natl. Acad. Sci. U. S. A. 77, 4216 – 4221. Van Tendeloo, V.F.I., Ponsaerts, P., Van Broekhoven, C., Berneman, Z.N., Van Bockstaele, D.R., 2000. Efficient generation of stably electrotransfected human hematopoetic cell lines without drug selection by consecutive facsorting. Cytometry 41, 31 – 35.
Recombinant Technology
A simple method for enriching populations of transfected CHO cells for cells of higher specific productivity S.C.G. Brezinsky, G.G. Chiang, A. Szilvasi, S. Mohan, R.I. Shapiro, A. MacLean, W. Sisk, G. Thill * Biogen Inc., 14 Cambridge Center, Cambridge, MA 02142, USA Received 9 July 2002; received in revised form 18 October 2002; accepted 15 January 2003
Abstract To establish a simple and rapid method for the screening of stable recombinant Chinese hamster ovary (CHO) cell lines, we have developed a cell surface labeling technique using fluorescently tagged antibodies that bind to secreted target proteins at low temperature. Using fluorescence intensity as the sole criterion for selection of cells, we are able to enrich populations of highly productive cells using preparative flow cytometry sorting. Reiterative sorting based on selection of cells having the highest fluorescence intensity of cell surface labeled protein results in dramatic increases in specific cellular productivity. Using lymphotoxin-beta receptor IgG fusion protein as a model system, we have demonstrated a greater than 20-fold increase in specific productivity (0.49 – 11.5 pg cell 1 day 1) (pcd) without the use of methotrexate (MTX)-mediated selection or amplification. In addition, the flow cytometry used to enrich for and clone high producer cell lines has reduced development time by more than 50% and the number of screening assays by more than 10-fold. When a transfected population of CHO cells expressing a humanized version of the murine monoclonal antibody (mAb) AQC2 directed against human alpha 1 beta 1 integrin was subjected to the same treatment, a 25-fold improvement in specific productivity (0.3 – 8.0 pcd) was observed. Furthermore, similar application of this technique to MTX-amplified clones resulted in up to 120-fold overall improvement in specific productivity (up to 42 pcd). Greater than 20 examples are also presented to demonstrate the robustness and performance of this technique. D 2003 Elsevier Science B.V. All rights reserved. Keywords: Recombinant protein production; High-speed cell sorting; Cell surface labeling; Rapid cell screening; Productivity enrichment; Flow cytometry
1. Introduction
Abbreviations: CHO, Chinese hamster ovary; LTbetaR, lymphotoxin beta receptor; MTX, methotrexate; GFP, green fluorescent protein; DHFR, dihydrofolate reductase; RPE, R-phycoerythrin; PI, propidium iodide; SPR, specific productivity rate. * Corresponding author. Tel.: +1-617-679-3275; fax: +1-617679-3148. E-mail address: [email protected] (G. Thill).
Flow cytometry has provided a fast, simple method for identifying and isolating rare cell phenotypes from complex populations. However, the normal release of a secreted protein from the plasma membrane has created a challenge for flow cytometry-based cell line development. Secreted proteins are synthesized on ribosomes bound to the endoplasmic reticulum (ER). Inside the ER, proteins achieve their native conformation via
0022-1759/03/$ - see front matter D 2003 Elsevier Science B.V. All rights reserved. doi:10.1016/S0022-1759(03)00108-X
142
S.C.G. Brezinsky et al. / Journal of Immunological Methods 277 (2003) 141–155
mechanisms involving disulfide bond formation and post-translational modifications such as glycosylation. The attainment of the correct conformation is the ratelimiting step in protein secretion. Subsequently, the protein is transferred to the Golgi apparatus and then to the plasma membrane where the secreted protein is released from secretory vesicles through the plasma membrane into the surrounding milieu. Some of the recent methods attempting to compensate for the resulting secretion-dependent loss of product-specific staining include immobilization of the secreted product near each cell by either encapsulation (Powell and Weaver, 1990; Gray et al., 1995) or a cellular affinity matrix (Holmes and Al-Rubeai, 1999). While these methods have greatly improved the ability to identify and isolate high producers, the immobilization protocols are often difficult, time consuming, and limited in the number of cells that can be examined at one time. Another promising technique uses co-selection of a nonsecreted surrogate marker like green fluorescent protein (GFP) (Meng et al., 2000). As the association of the protein with the cell surface is transient and downstream of the rate-limiting step in the secretory process, we hypothesized that the protein on the cell surface should be detectable and might correlate with the amount of protein being secreted from the cell. If so, then we should be able to label such cells with a fluorescent reagent that binds to the secreted product and sort the higher productivity cells using flow cytometry. Using this hypothesis, we have developed a very simple, fast method of lowtemperature direct staining of proteins on the cell surface followed by sorting, without the use of special media or long incubation times. High-speed cell sorters can sort hundreds of millions of cells with exceptional accuracy, greatly enriching high producer populations, and can deposit one cell per well into 96-well plates to facilitate cloning. Three rounds of reiterative sorting followed by single cell seeding has resulted in clones with specific productivity 20 times higher than the unsorted transfected population. A second cell line development project that involved selection of clones followed by methotrexate (MTX) amplification resulted in >100-fold enrichment in specific productivity up to a specific productivity of 42 pcd. Additionally, we present timelines, screening effort, and final titers for 23 recombinant cell lines to demonstrate the overall robustness of the technique.
2. Materials and methods 2.1. Vectors All plasmids utilize the CMV intermediate early promoter. The promoter extends from a restriction site approximately 500 bp upstream of the TATA box to a polylinker near the initiation codon of the native CMV intermediate early structural gene. The promoter region includes splice donor and acceptor sites in the 5Vuntranslated region. The polyadenylation site is derived from a human growth hormone variant sequence. Plasmid pXLTBR.9 contains the lymphotoxin-beta receptor gene fused to human IgG domains, CH2 and CH3 (LTbetaR-Ig) (Browning et al., 1995), driven by the CMV promoter together with a wild-type dhfr gene expression cassette for selection. Plasmids pAND162 and pAND160 are light and heavy chain vectors, respectively, used for the expression of AQC2, a humanized monoclonal antibody (mAb). Plasmid pAND162 contains a neomycin resistance selectable marker, while pAND160 contains the wild-type dhfr gene. 2.2. Cells, cell culture, and transfections Dihydrofolate reductase (DHFR) deficient DG44 Chinese hamster ovary (CHO) host cells (Urlaub and Chasin, 1980) were maintained as suspension cultures in serum-free medium containing nucleosides. Transfections were carried out by electroporation. Transfected AQC2 mAb cell lines were grown in alpha minus MEM supplemented with 10% dialyzed fetal bovine serum (FBS) (Hyclone Laboratories, Logan, UT) and 2 mM glutamine (Life Technologies, Grand Island, NY). The LTbetaR-Ig cell lines were grown in HYQPF-CHO, a commercial serum-free medium from Hyclone Laboratories, or serum-free alpha plus MEM medium (alpha plus MEMSF), an enriched alpha plus MEM without FBS. Following electroporation, the cells were cultured in six-well tissue culture dishes (Becton Dickinson Labware, Franklin Lakes, NJ). For the LTbetaR-Ig fusion protein cell line, DG44 host cells were transfected with pXLTBR.9 and cultured in alpha plus MEMSF. Twenty-four hours post-transfection, the growth medium was changed to HYQPF-CHO. Six to eight days post-transfection, the wells were pooled and split into three T-75 tissue culture flasks (Corning, Corning,
S.C.G. Brezinsky et al. / Journal of Immunological Methods 277 (2003) 141–155
NY). Each flask was then treated as an individual population and scanned by flow cytometry. Populations with flow cytometry profiles showing high fluorescent intensity subpopulations were subsequently subjected to high-speed cell sorting into growth media supplemented with antibiotic and antimycotic drugs (Life Technologies). For the AQC2 mAb cell line development, DG44 host cells were co-transfected with plasmids pAND162 and pAND160. Three days post-transfection, 400 Ag/ml G418 (Genetecin) (Life Technologies) was added to the medium. Once cells had reached f 80% confluence, the wells were pooled, fed with fresh media, and sorted within 24 h. As a rule of thumb, both serum-containing media and serum-free media have been used successfully from electroporation to clone selection. In most cases where serum is used, cells are adapted to growth in suspension culture in serum-free media after clone selection. 2.3. Surface fluorescent antibody staining of recombinant CHO for secreted product Cells ready for sorting were harvested by Accutase treatment (Innovative Cell Technologies, La Jolla, CA) and then maintained at 0– 4 jC for all subsequent handling. Accutase is a collagenase from a non-animal source that is preferred for its potential advantages from a regulatory standpoint. Trypsin is a suitable substitute from a technical standpoint. The cells were passed through a 70-Am nylon mesh (Becton Dickinson Labware), washed twice with cold phosphatebuffered saline (PBS) (Life Technologies), and then counted and assessed for viability. The cells were pelleted again by centrifugation for 5 min at 1000 RPM at 4 jC, and resuspended in cold DMEM/2% BSA containing fluorescently labeled antibody. In general, 5 106 to 5 107 cells are labeled with 50 Ag of R-phycoerythrin (RPE) conjugated goat F(abV)2 anti-human IgG (Jackson Immunoresearch Laboratories, West Grove, PA) in Dulbecco’s modified eagle medium (DMEM) (Life Technologies), supplemented with 2% bovine serum albumin (BSA) (Sigma, St. Louis, MO). RPE conjugates are preferred to FITC conjugates since their higher quantum yield of fluorescence lends itself to better signal-to-noise ratio. This is especially true during the initial sort where the percentage of positive populations is at its lowest
143
relative to subsequent sorts. After a 15-min incubation on ice, the cells were washed twice with cold PBS and resuspended in PBS plus 2 Ag/ml propidium iodide (PI) (Molecular Probes, Eugene, OR). Approximately 5 105 cells were removed for pre-analysis by FACSCalibur cytometer (see below) before sorting. DG44 host cells were used as a negative control. For photomicrography, cells were stained with Alexafluor 647 (Molecular Probes) following the same staining protocol described above. Alexafluor 647 is preferred in this application because of its resistance to photobleaching relative to other fluorescent conjugates. After staining, the cells were resuspended in PBS without PI. 2.4. Flow cytometry and fluorescent-activated cell sorting Analytical flow cytometry scans were performed on a FACSCalibur flow cytometer (Becton Dickinson, San Jose, CA) equipped with Cellquest v3.0 software and an air-cooled argon laser emitting at 488 nm. The PE emission was detected on Fl-2 using a 585/42-nmband pass filter, and PI emission on Fl-3, using a 670nm-long pass filter. Dead cells were excluded in an FSC vs. Fl-3 dot plot, and PE-labeled signal gating was done on a live cell-gated Fl-2 histogram. High-speed cell sorts were performed on a Moflo flow cytometer (Cytomation, Fort Collins, CO), equipped with Summit 3.0 software and an argon laser emitting at 488 nm for fluorescence excitation. The PE emission was detected on Fl-2, using a 580/30-nm-band pass filter, and the PI emission was detected on Fl-3 using a 670/30-band pass filter. Compensation of PE/PI emission spectrum overlap was accomplished using Cytomation’s Digital Signal Processing (DSP) board in Summit. Dead cells were excluded in an FSC vs. Fl-3 dot plot, and doublets were excluded in an FSC width vs. area dot plot. PElabeled signal gating was done on a live cell-gated Fl-2 histogram. The sorting gate was the combination of the live cell gate, the doublet discrimination gate, and the histogram gate on Fl-2. 2.5. Confocal microscopy Fluorescent and differential interference contrast (DIC) photomicrographs were acquired on a Leica TCS SP confocal microscope equipped with a red
144
S.C.G. Brezinsky et al. / Journal of Immunological Methods 277 (2003) 141–155
laser diode and Leica confocal software V2.00 build 0368 (Leica Microsystems, Heidelberg, Germany). Photomicrographs were taken of cells observed through a 40 -oil immersion objective. 2.6. Enzyme-linked immunosorbent assay (ELISA) LTbetaR-Ig titers were determined from media samples by ELISA. Assay plates were coated with an LtbetaR specific antibody (Biogen). Bound LTbetaR-Ig was detected with anti-human IgG horseradish peroxidase (HRP) conjugate (Jackson Immunoresearch Laboratories). For the quantitation of AQC2
mAb by ELISA, assay plates were coated with an AQC2-specific antigen fusion protein. Bound AQC2 mAb was detected with donkey anti-human IgG (H + L) horseradish peroxidase conjugate (Jackson Immunoresearch Laboratories). For both assays, the concentration was determined by backfit from a curve using four-parameter fit with SoftMax (Molecular Devices, Sunnyvale, CA). 2.7. Western blot analysis Protein samples and standards (1 and 5 Ag/ml) in a 10-Al volume were mixed with an equal volume of
Fig. 1. Confocal photomicrography showing product-specific cell surface staining of an AQC2-secreting CHO cell line. Both the AQC2 cell line (upper panels) and the untransfected CHO host DG44 (lower panels) were stained with Alexafluor 647-conjugated anti-human antibody. Fluorescent staining is shown on the left-hand side while differential interference contrast is shown on the right-hand side. The cells were kept on ice until microscopic analysis.
S.C.G. Brezinsky et al. / Journal of Immunological Methods 277 (2003) 141–155
2 SDS-PAGE sample buffer and boiled for 3 min. All samples were loaded on a 4 –20% SDS-PAGE gel (Biorad, Hercules, CA) and run at 40 mA (constant current) until the bromphenol blue reached the bottom
145
of the gel. Proteins were transferred to polyvinylidene fluoride (PVDF) membrane (Biorad) in 10 mM CAPS, 10% methanol, pH 11 overnight at 25 V at 4 jC. The membrane was blocked for 1 h in 5% BSA
Fig. 2. (A) Histogram of fluorescent-activated cell sort, negative control. The Fl-2 histogram was derived from the combination of the live cell gate based on PI exclusion (FSC vs. Fl-3), and the doublet discrimination gate (pulse width vs. FSC), to exclude doublets. (B) First LTbetaR-Ig fluorescent-activated cell sort, pre-sort. Histogram of red fluorescence on Fl-2. A sort gate R2 (not shown) was set to collect the brightest 5% of R-PE positive cells for all three reiterative sorts.
146
S.C.G. Brezinsky et al. / Journal of Immunological Methods 277 (2003) 141–155
in Tris-buffered saline with 0.05% Tween-20 (TBST). The primary antibody was an alkaline phosphatase conjugate of a goat F(abV)2 anti-mouse IgG (Jackson Immunoresearch Laboratories) used at 1/10 000 dilution in 2% BSA/TBST. The incubation was conducted for 1 h at room temperature with agitation. The filter was then washed three times at room temperature for 10 min in 0.5% BSA/TBST with agitation. The alkaline phosphatase reaction was conducted in 100 mM Tris, pH 10, 100 mM NaCl, 1 mM MgCl2, with a 1/200 dilution of 25 mM CDP-Star (Roche Molecular, Indianapolis, IN) for 5 min prior to exposure to film for 1 –4 min. 2.8. Determination of specific productivity rate (SPR) For each population or clone, 1 105 cells were seeded per well of a six-well tissue culture plate (Corning) in 2-ml growth media. Assays were performed in triplicate. The cells were allowed to grow for 3 days, conditioned media harvested for analysis, and the cells were removed by Accutase and counted. Specific LTbetaR-Ig and AQC2 antibody titers were quantitatively determined from media samples by ELISA. The SPR measured in picograms of specific protein per cell per day (pg cell 1 day 1: pcd) is a function of both growth rate and productivity, as represented by the following equations:
SPR ¼
ICA ¼
Total protein mass ¼ qP Integral cell area ðICAÞ
ðfinal cell number initial cell numberÞ days in culture loge ðfinal cell number=initial cell numberÞ
3. Results 3.1. Direct, product-specific staining of secreted recombinant proteins at the plasma membrane with fluorescently labeled antibodies Transfected cells secreting recombinant proteins can be directly stained with fluorescently labeled antibodies directed against the heterologous protein. An example of this is shown in Fig. 1. CHO cells secreting the humanized version of the antibody AQC2 were labeled with an Alexafluor 647 derivative of an anti-human antibody. The staining was then viewed using laser confocal microscopy. As seen in Fig. 1 (upper left quadrant), intense staining of the cell surface is demonstrated. The differential interference contrast (upper right) shows the specificity of the staining for the cell surface. Fig. 1 (lower panels) shows the results of the same staining regimen for the untransfected CHO host, DG44. No staining is seen in this case. These data confirm that the heterologous secreted protein can be directly stained on the cell surface without additional immobilization, and that this protein-specific staining can be used as the basis for flow cytometry sorting. 3.2. Generation of high-producing recombinant CHO cell lines in the absence of methotrexate LTbetaR-Ig: The CHO host DG44 was transfected by electroporation with the LTbetaR-Ig expression plasmid, pXLTBR.9. After selection and outgrowth of the transfectants for more than 14 days, the cells were pooled and labeled with an RPE-labeled F(abV)2 fragment of a goat anti-human IgG molecule at 4 jC. These cells, along with a stained untransfected host negative control, were subjected to analytical flow cytometry prior to a preparative sort. Both a live/dead
Fig. 3. (A) Analytical scan of a sample of LTbetaR-Ig cells collected after the first sort, showing that the sort resulted in a very pure population with an increase in mean fluorescence intensity and a corresponding increase in specific productivity. Approximately 2 – 3 105 cells were recovered. The SPR values were acquired after expansion of the cells in culture. (B) Analytical scan of a sample of collected LTbetaR-Ig cells after second sort, demonstrating progressive increases in both fluorescence intensity and specific productivity after reiterative sorting. Approximately 2 – 3 105 cells were recovered. The SPR values were acquired after expansion of the cells in culture. (C) Analytical scan of a sample of collected LTbetaR-Ig cells after third sort, showing no additional increase in either mean fluorescence or specific productivity over second sort. Approximately 2 – 3 105 cells were recovered. The SPR values were acquired after expansion of the cells in culture.
S.C.G. Brezinsky et al. / Journal of Immunological Methods 277 (2003) 141–155
147
148
S.C.G. Brezinsky et al. / Journal of Immunological Methods 277 (2003) 141–155
gate (forward scatter vs. Fl-3) and a gate to eliminate doublets (pulse width vs. FSC) were used to evaluate the signals from the negative control (Fig. 2A) and the sample. The sample (Fig. 2B) contained populations of cells from which the fluorescence intensity greatly exceeded that of the negative control. For preparative sorting, a gate was set which encompassed cells with the top 5% of the fluorescence intensity of the sample population and sorted cells subsequently collected. The cell number was then expanded by culture under selective conditions, and the process was repeated two more times. For all sorts of LTbetaR-Ig producing cell lines, an analytical scan was performed post-preparative sorting to insure the quality of the sort. The analytical scan and the experimentally determined mean specific productivity of the pool are displayed in Fig. 3A –C and Table 1. The unsorted transfected cell population had a specific productivity of 0.5 pcd. Sorting based on fluorescence intensity improved the specific productivity average of the pools by approximately a 10-fold improvement to 5.1 pcd (see Fig. 3C). The third sort as measured either by fluorescence intensity or ELISA assay and SPR calculation failed to demonstrate an increase in the specific productivity of the pool (Fig. 3B and C, Table 1). In fact, the mean fluorescence intensity (MFI) is lower relative to the second sort after the sorted cells from the third sort were subjected to analytical flow cytometry post-sort. Based on our experience, we ascribe no significance to this downward shift in MFI. It is related to variability in cytometer perform-
Table 1 Specific productivity rates (SPR) of unsorted and consecutively sorted CHO pools of LTbetaR-Ig demonstrate that specific productivity increases with reiterative sorting LTbetaR-Ig pool
Mean fluorescence intensity
Average SPR (pg cell 1 day 1) F S.D.
Unsorted First sort Second sort 2 Third sort 3
12.91 33.39 143.78 86.35
0.5 F 0.0 3.1 F 0.1 4.5 F 0.3 5.1 F 0.7
For the SPR assay, 1 105 cells were cultured in a six-well tissue culture dish in 2-ml serum-containing media. The cells and supernatant were harvested after 3 days in culture. The SPR assay was conducted in triplicate for each sample.
ance from run to run. This interpretation is supported by the SPR data that show a minimal increase in specific productivity. During the third sort, single cells were isolated directly from the cytometer into 96-well plates. The clonality of the seeded wells was verified by visual inspection. Several weeks later, clones were assayed for specific productivity and found to lie with a range of specific productivity between 3.6 and 11.5 pcd. The most productive clone represents a 23-fold enrichment in specific productivity without the need for MTX amplification. The number of clones assayed by ELISA during this enrichment was 50. The timeline was approximately 9 weeks from transfection to identification of the best clone. 3.3. Generation of methotrexate-amplified recombinant CHO cell lines AQC2 mAb: A second example of this technique is the expression of a humanized antibody to human alpha 1 beta 1 integrin. This approach underscores both the utility of this technique to enhance selection of high-producing clones and its usefulness in enriching and subcloning pools of MTX-amplified lines in order to meet stringent titer goals. The light and heavy chain were expressed from separate plasmids similar to that described for LTbetaR-Ig. After transfection and expansion under DHFR and G418 selection, the entire population, with a specific productivity of 0.3 pcd, was labeled with a fluorescent F(abV)2 fragment of goat anti-human IgG, and the top 2 –5% expressing cells as measured by fluorescence intensity were collected. After approximately 1 week of expansion, sorted cells were subjected to another sort. The cells were expanded again, then deposited at one cell per well into 96-well plates during the third sort, followed by visual verification of clonality. As in the case of LTbetaR-Ig, sorting resulted in a steady increase in the fluorescence intensity of the labeled cells as well as the measured specific productivity of both pools and clones (data not shown). Approximately 117 clones were expanded into 24-well plates and screened for antibody titer. The specific productivity of the highest expressing clones was determined in the SPR assay (Table 2). As the specific productivity of the best clones was not adequate for our needs, we performed a methotrexate
S.C.G. Brezinsky et al. / Journal of Immunological Methods 277 (2003) 141–155 Table 2 Specific productivity rates of AQC2 mAb CHO cells isolated from pools subjected to three reiterative FACS sorts before and after MTX amplification nM MTX Unamplified parent clone Amplified pool Amplified subclone Unamplified parent clone Amplified pool Amplified subclone
Amplified pool Amplified subclone
5A
SPR (pg cell 1 day1)
–
3.3
5AB 5AB-17 5AB-52 11B
250 250 250 –
16.6 41.0 32.3 8.0
11BB 11BB-46 11BB-67 11BB-68 11BB-83 11BA 11BA-1 11BA-30 11BA-41 11BA-47 11BA-50 11BA-118 11BA-100
250 250 250 250 250 100 100 100 100 100 100 100 100
13.5 25.4 27.3 19.9 26.9 17.9 18.4 19.9 18.2 26.6 26.1 28.1 32.5
No. of clones screened from amplified pool
52
121
126
Triplicate 3-day SPR assays were conducted in serum-containing media for each sample.
149
rescent-activated cell sorting and cloning, the best producers from 52 clones assayed exhibited qPs of 41.0 and 32.3 pcd. Similarly for clone 11B, cells amplified in 100 nM MTX had a mean specific productivity of 18 pcd and produced clones of up to 32.5 pcd out of approximately 126 clones screened (Table 2). The 250 nM MTX-amplified pool of clone 11B had a qP of 13.5 pcd and produced clones of up to 27 pcd in a similar size screen (Table 2). Fig. 4 shows a dramatic increase in the mean fluorescence intensity ( f 32-fold) of the top 100 nM MTX-amplified clone, 11BA-100, compared to the initial pool of transfectants, which correlates to the high-level product secretion ( f 110-fold increase in specific productivity). The increase in fluorescence intensity correlates to the dramatic increase in protein secretion. As was seen in the case of the LTbetaR-Ig fusion protein, fluorescence intensity is a very good surrogate marker for specific cellular productivity. This parameter can be used to guide cell line selection without handling individual clones during the selection process and without heavy reliance on ELISA assays. 3.4. Sorting improves per cell productivity
amplification on several clones. G418 was removed from the 10 highest expressing clones, prior to their being subjected to amplification in media containing either 100 or 250 nM MTX. Based on our experience, we favor methotrexate concentrations in the 25 –250 nM range as they are less likely than higher concentrations to slow the growth rate of the production cell line but can still generate specific productivity increases. Amplified pools were screened for antibody titer. Populations exhibiting a qP equal to or greater than 13.5 pcd were subjected to highspeed cell sorting and autocloning of the highest 2% expressing population into 96-well plates. Subcloning is required at this point because methotrexate amplification is a stochastic process generating heterogeneity in the treated population. Two of the top antibody-producing clones, clone 5A and 11B, had unamplified qPs of 3.3 and 8.0 pcd, respectively (Table 2). When clone 5A was amplified in the presence of 250 nM MTX, a pool of specific productivity of 16.6 pcd was generated. After fluo-
As a control experiment to demonstrate the utility of cell surface labeling and flow cytometry sorting in enriching populations of cells for high producers, we compared the following two samples. The first is a non-amplified transfection expressing a human cytokine receptor fused to human Fc (Table 4, line 3) sorted multiple times. After cloning, 94 out of 106 (89%) of clones sorted two or three times produced greater than 1 Ag/ml in an assay of 24-h old media supernatants. We found that 45% of the positive clones produced >5 Ag/ml and 9% produced >10 Ag/ml. Table 3 shows the distribution of ELISA titers in this experiment. When a collection of randomly selected clones from a similar transfection treated with 25 nM MTX to boost productivity was similarly assayed, only 8 out of 100 (8%) were positive by the same criterion. All clones produced between 1 and 5 Ag/ml. A second control experiment was performed as follows. A cell line was identified (Table 4, line 18) that had been shown to produce Western blot positive
150
S.C.G. Brezinsky et al. / Journal of Immunological Methods 277 (2003) 141–155
Fig. 4. Analytical scan of unsorted AQC2 mAb CHO pool (black) vs. amplified clone 11BA-100 (green) demonstrates an increase in both mean fluorescence intensity and specific productivity after MTX amplification, sorting, and cloning. The Fl-2 histograms were derived by analysis of the PE signal within the live cell gate.
clones (>1 Ag/ml by visual inspection in 8 out of 8 cases and >5 Ag/ml in 7 out of 8 cases). In order to demonstrate the magnitude of the enrichment produced by cell surface labeling and sorting, this line was resorted from a frozen vial of previously unsorted
Table 3 Enrichment of clones for titer using fluorescent labeling and flow cytometry Criterion (Ag/ml)
Random clone selection
Two to three enrichment sorts
EXAMPLE #1a <1 1–5 5 – 10 >10
92/100 8/100 0 0
12/106 54/106 32/106 8/106
EXAMPLE #2b <1 1–5 >5
24/26 2/26 0
0 8/8 7/8
a Samples were measured by ELISA after a 24-h incubation in fresh media. b Samples were measured after a 24-h incubation in fresh media in a Western blot format and assessed by visual inspection.
cells from the same transfection. Following labeling, cells were sorted into 96-well plates with the Fl-2 sort gate set at 100%, i.e., without enrichment. Following outgrowth, media conditioned for 24 h from nearly confluent plates was assayed by Western blot and compared to a 1 Ag/ml standard. We found that 2 out of 26 lines analyzed (8%) met the 1 Ag/ml criterion and barely so (data not shown). Upon prolonged exposure, the protein could be seen in about half of the samples, but was very much reduced in intensity relative to the 1 Ag/ml standard. In both cases, sorting improved both the percentage of positive clones and the titers produced by the positive clones relative to unsorted cells (Table 3). 3.5. Historical list of applications Table 4 lists a series of 23 cell lines developed using this method. A generic description of the protein is noted in the first column of the table. The Fc portion of all the receptor-Fc fusion proteins is either human IgG1 or murine IgG2b. Two examples, lines 5 and 16, which were not receptor-Fc fusion proteins, used murine monoclonal antibodies followed by an RPE-
S.C.G. Brezinsky et al. / Journal of Immunological Methods 277 (2003) 141–155 Table 4 Results obtained from multiple cell lines using flow cytometry to enrich for increased per cell productivity Product
No. of Final No. of Timeline sorts titer assays to clone mg/la ID (weeks)
(1) (2) (3) (4)
3 2 3 2
150 140 200 180
100 20 106 20
12 7 7 8
2
50
11
7
2
40
24
7
2
100
20
7
2
50
20
8
2
33
10
7
2
70
10
6.5
2 2 2 2 3 2
125 55 100 75 40 75
10 10 20 10 150 20
7 7.5 8 7 9 9
2 2 2
60 160 70
20 8 20
6 7 6.5
3 2
166 130
148 18
8 8
2 3
150 242
18 129
7 8
Immune system ReD/hFc Type III ReD/hFc Hormone ReD/hFc Secreted version of GPI-linked protein/FC (5) Murine adhesion molecule/Flag tag (6) Murine immune system ReD/hFc mutantA (7) Murine immune system ReD/hFc mutantB (8) Human immune system ReD/hFc mutantA (9) Human immune system ReD/hFc mutantB (10) Human immune system ReD/hFc mutantC (11) Human ReD/Fc (12) Mutant of ReD/Fc A (13) Mutant of ReD/Fc B (14) Mutant of ReD/Fc (15) Mutant of ReD/Fc (16) Secreted version of GPI-linked protein/no tag (17) Cytokine ReD/Fc (18) Murine ReD/MigG2b (19) Murine kidney surface molecule-hFC (20) Antibody (21) Murine immune system ReD/mFc (22) MReD-hFc (23) Antibody #2
Receptor extracellular domain (ReD). a Cells were adapted to growth in suspension in serum-free media. Titer was determined by A280 following protein A capture in most cases.
labeled anti-mouse conjugate for detection. Cell lines used for preliminary research experiments, e.g. in vitro experiments or small animal disease models, are subjected to two sorts and a small number of assays are performed. Western blots of media conditioned for 24 h are measured relative to standards (1 and 5 Ag/ml). This assay is sufficient to permit the choice of a line capable of generating enough protein for drug candidate evaluation studies. Cells destined for commercial purposes are sorted and assayed more extensively as
151
higher titer may be necessary. In most cases, 6 to 8 weeks is all that is needed to identify a number of clones capable of producing titers of 50– 200 mg/l. Methotrexate is not generally used in the construction of these lines as it requires additional time to amplify and select clones. The only difficult protein on the list is #15 which is the last in a series of mutants (11 –15) and the only one that did not easily yield good clones. This protein seemed to be adhering to the cell surface, i.e., was FACS positive but without secretion into the medium in almost all cases. However, when a positive clone was found, it was amplified and clones producing >150 mg/l were easily found. We have noted failures using this approach related to ‘‘protein engineering’’ problems, e.g. attempts to pentamerize antibodies (three examples). We also had two examples where the protein was toxic (a large amount of cell death accompanied both the initial transfection and subsequent enrichment steps) to cells and did not yield productive clones. In addition, the use of commercial vectors, e.g. pCMV Script (Stratagene, La Jolla, CA) is not advisable. This technique is very sensitive to the ability of the cell to deliver the desired protein to the cell surface at a high enough level to produce a signal. Transcriptionally weak vectors do not produce enough protein to do so. We strongly recommend that only vectors such as the one described herein be used for the purposes of flow cytometry-based cell line development. 3.6. Other applications We have investigated the applicability of this technique to other cell types that produce proteins of interest to the biotechnology field. We labeled both a hybridoma (murine monoclonal IgG1) against a human cell surface protein and its fusion partner with an RPE-labeled conjugate directed against heavy and light chains. Results are shown in Fig. 5A. The black and green curves show the fusion partner in the absence and presence of staining. Mean fluorescence intensities are comparable. The cultured hybridoma reacted strongly with the antimouse IgG conjugate (red curve) and had a mean fluorescence intensity almost 60-fold greater than background. It also appears to contain at least two populations of cells, one a low abundance high-
152
S.C.G. Brezinsky et al. / Journal of Immunological Methods 277 (2003) 141–155
Fig. 5. Histograms of red fluorescence intensity. There were 1 – 2 106 cells labeled with 2 Ag RPE-labeled goat anti-mouse IgG (H + L). (A) FL653 fusion partner unstained (black), FL653 fusion partner stained (green), A8 G3.5 monoclonal antibody (red). (B) NS/0 host unstained (black), NS/0 host stained (green), mouse anti-human CD154 antibody (red).
expressing group and a more abundant population with a much lower MFI. As a second example of alternative hosts, we examined the expression of a murine anti-human CD154 antibody from transfected NS/0 cells using the cell surface staining technique. The nontransfected host (Fig. 5B) had an MFI of 1.6, while the stained version of the host was about twofold higher. This indicates a background level of antibody or light chain secretion from or cell surface localization on the host. The transfected host expressing the antibody had an
MFI greater than 30-fold showing that it was clearly positive. The most surprising result though was the breadth peak of MFI in the histogram. We do not see such broad peaks for clonal lines in CHO cells.
4. Discussion The screening method described herein, for transfected cell lines producing heterologous proteins, relies on the fact that the secreted protein is transiently
S.C.G. Brezinsky et al. / Journal of Immunological Methods 277 (2003) 141–155
associated with the cell surface during protein secretion. As such, the cell can be labeled with a fluorescent reagent such as a protein-specific antibody. Fluorescently labeled cells are excellent substrates for flow cytometry. A study by Kromenaker and Srienc (1994) involving direct staining and sorting of hybridomas indicated that fluorescence of surfaceassociated IgG is stable for an extended period of time. These data correlate with our own findings that the fluorescence of directly IgG-stained CHO producer cells is stable for over 1 h at room temperature as measured by flow cytometry and laser confocal microscopy (unpublished observations). The effectiveness of this method is demonstrated by the fact that, in the two cases presented in detail in this paper, fluorescence intensity of cell surface proteins was used as the predominant criteria for selection of clones and resulted in the selection of clones with relatively high-specific productivity. A similar direct relationship has been found between surface IgG expression and antibody secretion rate in hybridomas, although this is not true for all cell lines (Sen et al., 1990; Cherlet et al., 1995; Marder et al., 1990; McKinney et al., 1991). The technique has several advantages from a handling standpoint. First, it is thorough since all cells present in the transfected or amplified populations are examined. Second, all cells up to the point of cloning are handled in batch. As very little handling is involved, it is possible for a single person to produce multiple cell lines. Immunoassay labor is also greatly reduced as initial screening is performed by the cytometer. Relative to techniques such as Gel Microdrop technology, this technique is simpler and more reliable (unpublished observations). In addition, at over 100 million cells sorted after a single staining event, we have not yet reached the limit on the number of cells capable of being processed with this protocol (unpublished observations). In addition, the continuing improvements in fluorescent-activated cell sorting technology have improved the accuracy of population enrichment and single cell seeding. The current methodology directly measures the presence of the secreted heterologous protein on the cell surface and is potentially superior to surrogate markers such as co-transfected green fluorescence protein (GFP). GFP is a good measure of transcription, but secretion of heterologous proteins is not
153
necessarily a linear process. Overexpression of protein can deplete oxidizing reagents involved in proper folding of proteins leading to misfolding and inefficient secretion (Parekh and Wittrup, 1997). Hence, increased transcription does not necessarily correlate with increased secretion. Our method selects for the cell line that possesses the proper cellular machinery and conditions (integration site or MTX concentration, etc.) that lead to maximal presence on the cell surface, which is a better predictor of secretion. The two described examples are representative of many of applications of this technique (see Table 4). In the first case, an unamplified line was generated in approximately 9 weeks from transfection to identification of the best clone. The timeline for MTX amplification about twice as long as sorting is repeated after several clones are subjected to multiple MTX concentrations. This fluorescent-activated cell sorting method is also a good analytical method for screening several clones under multiple MTX concentrations and choosing the best combination of both. As shown in Table 4, a large number of cell lines have been developed using this technique. These lines were developed in a relatively short period of time. Mean development time is 7.6 weeks for clone selection. Average assay number is 43 assays per line. Greater than 50% of the clones produced the protein at >100 mg/l. We have performed two control experiments to show that cell surface labeling followed by sorting enriches the number of positive clones for a given expression criteria. In both experiments, the increase in number of positives was >10-fold. In addition, among the positive clones generated by this method, the average titer was significantly higher (see Table 3). Preliminary experiments using this technique on hybridoma lines and transfected NS/0 cells indicate that the method can detect the presence of secreted antibody in these hosts as well as in CHO transfectants. This technique may be useful in hybridoma subcloning, isotype selection, or in monitoring the stability of expression during routine passage. The hybridoma line had two separate populations, the less abundant having the higher MFI. This may be a reflection of selection pressure against the high producer population mediated by the genetic instability of the myeloma line. A second phenomenon of note was the small increment of background staining in the NS/0 line. Myeloma lines are
154
S.C.G. Brezinsky et al. / Journal of Immunological Methods 277 (2003) 141–155
professional antibody producing cells which, coupled with their unstable genome, could lead to spontaneous antibody production. Such background signals would clearly interfere with selection of high-producing clones. Myeloma lines are qualitatively different from CHO cells being lymphoid cells as opposed to the CHO epithelial lineage. This may impact the nature of the glycocalyx (see discussion below on mechanism) and require a more thorough investigation of cell culture and fluorescent labeling conditions in order to optimize this technique, Nevertheless, at first approximation, this technique seems to be capable of detecting antibody secretion in both hosts. Although we have not thoroughly explored the mechanism by which the secreted protein is held onto the cell surface and easily detected, we believe that there are two leading hypotheses: (1) staining of the secreted protein on the plasma membrane as it is being released from secretory vesicles or (2) transient association of the secreted protein with the protein or proteoglycan components of the glycocalyx (Bennett,1963). We favor the glycocalyx hypothesis for several reasons. The first relates to stability of signal, which is greater than 1 h. During secretion, the protein is packaged into secretory vesicles and rapidly transported to the plasma membrane. Vesicles fuse on the cytoplasmic side of the membrane, and the contents of the secretory vesicle cross the membrane through an aqueous pore (Lippincott-Schwartz et al., 2000; Scales et al., 2000). We have investigated the temperature dependence of labeling to see if labeling the cells at 4 jC, due to perturbation of membrane fluidity, is a component of signal generation and strength. Comparable signals were obtained at 4, 23, and 37 jC (unpublished observations). Given the temperature independence of cell surface labeling coupled with the rapidity of secretory vesicle fusion to the plasma membrane, and in contrast to the longevity of the signal, we doubt that we are observing labeling of the protein on the plasma membrane. Second, there is a precedence for binding of secreted proteins to the glycocalyx. This phenomenon of cell surface attachment for secreted proteins has been observed for FGF-2 (Engling et al., 2002). FGF2 is delivered to the cell surface via a nonclassical secretion mechanism and is found attached to surface proteoglycan.
Third, we believe that the glycocalyx hypothesis explains one of the more puzzling aspect of the technology, false positives. Every cell line development project encounters some percentage of false positives. False positives are clones with high mean fluorescence intensity values but no protein secretion. Table 4, lines 11 –15, shows a series of mutants of an immune system receptor-Fc fusion protein which posed no problems in cell line development save for cell line #15. This molecule showed an unusually high percentage of false positives (>95%). When some of these nonsecretors were reinterrogated for fluorescence intensity, MFI was well above background. Our explanation was that the protein was produced at a low level and sticking to the cell surface producing a fluorescent signal but not being secreted. Only two clones secreting the protein were ever found out of f 150 screened. One of these clones was capable of being amplified by MTX (50 nM) treatment. It produced high levels of protein in 14 out of 14 clones surveyed. These observations are consistent with proteins being localized to the cell surface via interaction with the cell surface glycoproteins. Once a binding threshold was exceeded, secretion was facile. These observations are not easily explained with a plasma membrane/secretory vesicle model, especially when considering the saturation aspects of the phenomenon. Lastly, the stability of the sorted populations, as reflected in change in mean fluorescence intensity between sorts, is a good indicator of the physiological impact of expression of a given heterologous protein on the host. The physiological impact ranges from excellent stability to severe collapse of the mean fluorescence intensity. The latter case may reflect toxicity of the foreign protein to the host, differential growth rates between cells with different expression levels, or slow adaptation of high producers when cultured at low density. Knowledge of these phenomena can assist in planning screening strategies (number of sorts, number of clones to assay, choice of slow vs. fast growers).
Acknowledgements The authors acknowledge the support of Konrad Miatkowski during the development of this technology. We thank Stephan Miller for guidance and
S.C.G. Brezinsky et al. / Journal of Immunological Methods 277 (2003) 141–155
assistance in manuscript preparation. We are also grateful to John McCoy for critical comments on the content of the manuscript.
References Bennett, H.S., 1963. Morphological aspects of extracellular polysaccharides. J. Histochem. Cytochem. 11, 14 – 23. Borth, N., Zeyda, M., Katinger, H., 2000. Efficient subcloning of high-producing subclones during gene amplification of Chinese hamster ovary cells by flow cytometry and cell sorting. Biotechnol. Bioeng. 71, 266 – 273. Browning, J.L., Dougas, I., Ngam-Ek, A., Bourdon, P.R., Ehrenfels, B.N., Miatkowski, K., Zafari, M., Yampaglia, A.M., Lawton, P., Meier, W., Benjamin, C.P., Hession, C., 1995. Characterization of surface lymphotoxin forms. Use of monoclonal antibodies and soluble receptors. J. Immunol. 154, 33. Cherlet, M., Kromenaker, S.J., Sirenc, F., 1995. Surface IgG content of murine hybridomas: direct evidence for variation of antibody secretion rates during the cell cycle. Biotechnol. Bioeng. 47, 535 – 540. Engling, A., Backhaus, R., Stegmayer, C., Zehe, C., Selenmeyer, C., Kehlenbach, A., Schwappach, B., Wegehingel, S., Nickel, W., 2002. Biosynthetic FGF-2 is targeted to non-lipid rafts following translocation to extracellular surface of CHO cells. J. Cell Sci. 115, 3619 – 3631. Gray, F., Kenny, J.S., Dunne, J.F., 1995. Secretion capture and report web: use of affinity derivatized agarose microdroplets for selection of hybridoma cells. J. Immunol. Methods 182, 155 – 163. Holmes, P., Al-Rubeai, M., 1999. Improved cell line development by a high throughput affinity capture surface display technique to select for high secretors. J. Immunol. Methods 230, 141 – 147. Kaufman, R.J., Sharpe, P.A., 1982. Amplification and expression of sequences cotransfected with a modular dihydrofolate reductase complementary DNA gene. J. Mol. Biol. 159, 601 – 621. Kim, N.S., Byun, T.H., Lee, G.M., 2001. Key determinants in the occurrence of clonal variation in humanized antibody expression
155
of CHO cells during dihydrofolate reductase mediated gene amplification. Biotechnol. Prog. 17, 69 – 75. Kromenaker, S.J., Srienc, F., 1994. Stability of producer hybridoma cell lines after cell sorting: a case study. Biotechnol. Prog. 10, 299 – 307. Lippincott-Schwartz, J., Roberts, T.H., Hirschberg, K., 2000. Secretory protein trafficking and organelle dynamics in living cells. Annu. Rev. Cell Dev. Biol. 16, 557 – 589. Marder, P., Maciak, R.S., Fouts, R.L., Baker, R.S., Starling, J.L., 1990. Selective cloning of hybridoma cells for enhanced immunoglobulin production using flow cytometric cell sorting and automated laser nephelometry. Cytometry 11, 498 – 505. McKinney, K.L., Dilwith, R., Belfort, G., 1991. Manipulation of heterogeneous hybridoma cultures for overproduction of monoclonal antibodies. Biotechnol. Prog. 7, 445 – 454. McKinney, K.L., Dilwith, R., Belfort, G., 1995. Optimizing antibody production in batch hybridoma culture. J. Biotechnol. 40, 31 – 48. Meng, Y.G., Liang, J., Wong, W.L., Chisholm., V., 2000. Green fluorescent protein as a second selectable marker for selection of high producing clones from transfected CHO cells. Gene 25;242 (1 – 2), 201 – 207. Parekh, R.N., Wittrup, K.D., 1997. Expression level tuning for optimal heterologous protein secretion in Saccharomyces cerevisiae. Biotechnol. Prog. 13 (2), 117 – 122. Powell, K.T., Weaver, J.C., 1990. Gel microdroplets and flow cytometry: rapid determination of antibody secretion by individual cells within a cell population. Bio/Technology 8, 333 – 337. Scales, S.J., Bock, J.B., Scheller, R.H., 2000. The specifics of membrane fusion. Nature 407, 144 – 146. Sen, S., Hu, W.-S., Srienc, F., 1990. Flow cytometric study of hybridoma cell culture: correlation between cell surface fluorescence and IgG production rate. Enzyme Microb. Technol. 12, 571 – 577. Urlaub, G., Chasin, L., 1980. Isolation of Chinese hamster ovary cell mutants deficient in dihydrofolate reductase activity. Proc. Natl. Acad. Sci. U. S. A. 77, 4216 – 4221. Van Tendeloo, V.F.I., Ponsaerts, P., Van Broekhoven, C., Berneman, Z.N., Van Bockstaele, D.R., 2000. Efficient generation of stably electrotransfected human hematopoetic cell lines without drug selection by consecutive facsorting. Cytometry 41, 31 – 35.