Concentrating Feed—an Applicable Approach to Improve Antibody Production

Recombinant protein production in the biopharmaceutical industry mostly relies on fed-batch mammalian cell culture (1, 2). Often in fed-batch processes, the feed volume can be more than 1000 L for a 5000-L bioreactor with 30 to 40% (v/v) feed. To alleviate workload in media preparation and simplify the manufacturing process, a reduction in feed volume is greatly desired. Feed-volume reduction can also reduce the cost of consumables, such as disposable bags and medium storage space. Because the feed dilutes the culture and reduces product concentration, the use of lower feed volumes can improve overall process productivity as well.

One approach of decreasing feed volumes but still stoichiometrically supplying nutrients to cells is to concentrate the feed media. The preparation of highly concentrated feed media is, however, challenging because of the limited solubility of medium components. The solubility of many amino acids can be increased by adjusting the pH (3). Recently, surfactants have been used to increase the solubility of chemicals in the feed media (4). In this study, the authors compared strategies including raising the pH and adding surfactant(s) in the preparation of concentrated feed media. In the experiment, the authors fed less volume of concentrated feeds to the cell cultures, as opposed to the constant volume feeding strategy used by Hossler et al. (4). Among the surfactants evaluated, Poloxamer 188 (P188) (i.e., Pluronic F68) is a nonionic surfactant that is an additive in mammalian cell culture media and drug substance formulation (5-7). It was found that the feed volume could be reduced by 2.5-fold with the concentrated media supplemented with higher levels of P188, while volumetric productivity was improved by 10-50%. The improvement of productivities was likely due to the enhancement of protein folding and secretion machinery suggested by quantitative polymerase chain reaction (qPCR) array assay. Supplementing surfactant and raising pH approaches were successfully used in the preparation of concentrated feed media for 7-L and 5000-L bioreactor runs, respectively. These straightforward approaches can be applied to large-scale production of recombinant proteins.

Materials and methodsCell lines. Three recombinant Chinese hamster ovary (CHO) cell lines, developed in-house to produce three different monoclonal antibodies, were used in this study. Cell lines A and B are dihydrofolate reductase-deficient (DHFR-deficient) DG44 cell lines. Cell line C is a CHO glutamine synthetase (GS) cell line. Frozen vials of the cell lines were thawed and grown in shake flasks containing Bristol-Myers Squibb’s proprietary basal media. The flasks were shaken in incubators at 36.5 °C with 5% CO₂.

Media preparation. To prepare feed media for cell lines A and B, the formulated powder was added into water for injection (WFI) or cell-culture-grade water at 37°C, followed by addition of Polysorbate 80 (PS80) or P188. 10 N sodium hydroxide (NaOH) was used to raise the pH until the solution was clear. After the chemicals were completely dissolved, the medium was made up to the required volume by adding WFI or cell-culture-grade water and filtered through a 0.22-µm filter (Millipore). The medium pH was measured by a pH meter (Thermo Scientific), and the medium osmolality was measured by an osmometer (Advanced Instruments). This medium is referenced as M1 in Table I.

To prepare the feed medium for cell line C, the formulated powders were added into WFI or cell-culture-grade water sequentially at room temperature. 10 N NaOH was used to raise the pH until the solution was clear. After mixing well, the solution was made up to the required volume and filtered through a 0.22 µm filter (Millipore). The pH and osmolality of the medium were measured. This medium is referenced as M2 in Table I.

Fed-batch studies in shake flasks and bioreactors. Fed-batch studies in shake flasks were carried out in 250-mL Erlenmeyer flasks in incubators at 36.5 oC with 5% CO2. Applikon 7-L benchtop bioreactors were used for scaling up shake-flask studies. The process for cell line C was scaled up to the 5000-L production bioreactor.

Cell counts, antibody titer, protein quality analysis, and qPCR array. Cell-culture samples were counted by a Vi-CELL (Beckman-Coulter) using the trypan blue exclusion method. The pH and metabolites in spent media were measured by a NOVA FLEX instrument (Nova Biomedical). Antibody titer was measured using high-performance liquid chromatography (HPLC) (Waters Corporation) equipped with a protein A column (Applied Biosystems). The protein charge profile was analyzed by imaged capillary isoelectric focusing (iCIEF) (ProteinSimple, Santa Clara, California). Antibody N-glycosylated forms were analyzed by PNGase F digestion, reductive amination labeling, followed by separation using ultra-performance liquid chromatography (UPLC).

In a qPCR array, 3 × 10⁶ cells of day 9 and day 11 from each treatment were used to extract messenger ribonucleic acid (mRNA) following the protocol of RNeasy Mini Kit (Qiagen). 400 ng of total RNA for each sample was reverse transcribed into complementary deoxyribonucleic acid (cDNA) using RT2 First Strand Kit (Qiagen). Four experimental cocktail solutions composed of cDNA synthesis reaction and SABiosciences RT2 qPCR master Mix (Qiagen) were dispensed into 4 × 96 wells (10 µL/well) in 384-well PCR unfolded protein response array (Qiagen). The array was analyzed by ViiA 7 real-time PCR system (Life Technologies). The raw CT values were further analyzed by free web-based RT2 profiler PCR array data analysis version 3.5 (Qiagen). The ratios of gene levels between day 9 and day 11 for each treatment were reported. The gene functions were obtained from the website of the National Center for Biotechnology Information (www.ncbi.nlm.nih.gov/gene).

ResultsConcentrated feed medium M1 test in shake flasks. In the concentrated feed medium study, the original concentration is designated as 1X. The 1X medium was used as a control. Feed media with concentrations of 1X, 1.30X, 1.60X, 1.75X, 1.90X, and 2.50X were prepared by adding 10 N NaOH to increase the pH and dissolve all of the components into the solutions. The higher the feed concentration, the more the base needed for solubility, resulting in an increase of pH from pH 8.58 for 1X medium to pH 10.41 for 2.50X medium, as shown in Table I. The osmolality of the concentrated feed media was increased from 835 mOsm/kg of 1X medium to 2432 mOsm/kg for 2.50X medium. To test if a surfactant can increase the solubility of medium components without increasing the medium pH, 2.50X concentrated feed media with various percentages of PS80 or higher levels of P188 were made. The data show that these media required less NaOH compared to the 2.50X concentrated medium, resulting in a lower final medium pH and lower osmolality (see Table I).

The concentrated feed media described previously were used in the fed-batch culture of cell line A. The feed volumes are shown in Table I. Nutrients were added to the cell culture for the concentrated feed media at the same quantities for the 1X feed medium. The productivities on the last day of the process (day 14) showed that the 1.60X medium led to 12% and 20% increase of the titer and specific productivity respectively, which is significantly higher than the 1X medium control (p < 0.05). The 2.50X medium had the lowest productivity (see Figure 1a), suggesting that there was an upper concentration limit of feed media prepared by raising the pH, which was due to the lower cell viability at higher cell culture pH and osmolality (data not shown). Among the surfactants evaluated, the addition of 0.5% (w/v) P188 in 2.50X medium increased the titer and specific productivity by 15% and 25%, respectively, while the feed volume could be reduced by 2.5-fold. More favorable culture pH and osmolality profiles from the 2.50X + 0.5% P188 feed medium have been observed as these are closer to the 1X control condition (data not shown). The 2.50X + 0.5% P188 feed medium also led to 42% and 116% increase in the final titer and specific productivity of cell line B respectively (see Figure 1b).

Antibody production in 7-L bioreactors using concentrated feed media of M1. To test the scalability as well as to remove the negative effects from the high pH feed media, 7-L bioreactors were used to run cell line A process with pH control. The profiles of viable cell density and viability are shown in Figure 2a and 2b. The 2.50X + 0.5% P188 medium was tested in parallel to the 1X medium that was used as a control. The viability for the control decreased from day 8, whereas the concentrated feed maintained higher viability for a longer period of time. The productivity profiles shown in Figure 2c indicate that use of the concentrated feed medium improved the productivity by 33% towards the end of the process. The use of the concentrated feed medium also led to a significantly higher (p < 0.05) specific productivity than the control (data not shown). These results suggest that productivity can be improved when feed volume is reduced, which is consistent with the shake-flask studies. The product qualities, including charge heterogeneity and N-glycan profiling, were not affected by the increase of P188 concentration in the feed (see Figure 3). There were no changes in antibody monomer percentages and protein purity with addition of P188 (data not shown).

Antibody production in 5000-L bioreactor using concentrated feed media of M2. The concept of concentrated media was applied in the development of a feed medium for the process of cell line C. A feed medium was developed by concentrating the original feed medium to 7.5 fold, designated as M2 (see Table I). From M2, a 10-fold increase of the original feed medium was achieved by adjusting pH to between 8.8 and 9.5, leading to 1.33X M2 feed medium. The two media were tested in parallel using 7-L bioreactors. The feeding volume of 1.33X M2 was reduced by 33% so that the total amount of each component added to the bioreactor was the same as that of the 1X M2 control. Data from a 5000-L manufacturing bioreactor run performed using 1.33X M2 feed medium are overlaid in Figure 2d. The productivity of 1.33X M2 was 11% higher than that of 1X M2 at both development and 5000-L manufacturing scales.

Gene regulation by feeding with concentrated media. To understand the rationale for improving antibody specific productivities in cultures fed with concentrated feed media (see Figure 1), an unfolded protein response qPCR array was chosen in this study. The array included 84 genes for protein expression, folding, and secretion. Gene level changes from day 9 to day 11 in the protein-production phase were investigated in this assay, using two samples prepared on both days from 1X M1 feed medium treatment and two samples from 2.50X + 0.5% P188 treatment. Gene levels that changed more than 1.5-fold from these treatments are listed in Table II. The downregulated genes in the control samples (highlighted in blue) include heat shock protein family genes (Hspa5, Hsp90b1), Ca2+-binding chaperones (Calr3, Calr), protein disulfide isomerase family gene (Pdia3), nucleotide exchange factor for unfolded proteins (Sil1), ubiquitin-like domain protein for endoplasmic reticulum (ER)-associated degradation (Herpud1), and three unknown genes. Instead of downregulation, these genes were either upregulated or non-differentially expressed in cells fed with 2.50X + 0.5% P188 (see Table II). Highlighted in gray in Table II, Ero1 encoding an ER oxidoreductase and Hspa1L encoding an ER chaperone were non-differentially expressed in cells fed with the 1X medium, but up-regulated in cells fed with 2.50X + 0.5% P188. These data suggest that the 2.50X + 0.5% P188 medium could maintain and elevate levels of genes for chaperones and folding enzymes that ensure protein quality control in the secretory pathway. This finding is consistent with the higher cell viability and higher productivity observed in the concentrated medium.

Among the up-regulated genes of the 1X M1 medium treatment (highlighted in orange in Table II), Srebf1, Insig1, and Scap are genes for sterol biosynthesis (8). These genes were non-differentially expressed in cells fed with 2.50X + 0.5% P188. P188, a surfactant consisting of hydrophobic and hydrophilic blocks, has been reported to reduce cell death (9, 10). It may play a role similar to lipids, which have hydrophobic tails and hydrophilic heads that confer cell protection from stress. In addition to three sterol synthesis genes, Htra3, a serine peptidase gene, was upregulated more than four-fold from day 9 to day 11 in the 1X M1 medium, while only about two-fold in the concentrated medium. Given that serine peptidase contributes to protein degradation, a higher level of this protease may be one of the reasons for the lower antibody levels observed in the 1X M1 medium compared to that in the 2.50X + 0.5% P188 medium.

Discussion
In this study, different strategies of preparing concentrated feed media were compared. Although higher concentrations of feed media could be obtained by raising pH, there was an optimal pH of feed media in which cell growth and antibody production were not adversely affected. The addition of a surfactant makes it possible to prepare more concentrated feed media while maintaining the pH. In this study, 2.50X concentrated medium supplemented with 0.5% P188 led to a 10-50% increase in titers while feeding as low as 40% of the original feed volume. The surfactant addition to the feed did not have any impact on measured protein quality attributes.

Medium storage stability is a general concern for manufacturing and development using concentrated feed media. In this study, the 2.50X + 0.5% P188 M1 and 1.33X M2 media have a shelf life of 60 days at 2-4 oC and 26 days at ambient temperature respectively (data not shown). There was no precipitation observed and antibody titers were not affected. These data show that concentrating feed medium either by raising pH or adding surfactant is an effective and applicable approach in simplifying the manufacturing process and improving productivity.

Acknowledgements
The authors would like to thank Gautam Nayar, Joel Goldstein, and Frank Ritacco for their review and comments on this manuscript. The authors would also like to thank Amanda Bell, Wenkui Lan, and Ya Fu for their support in protein purification and protein quality analysis.

References
1. B. Kelley, mAbs. 1 (5) 443-452 (2009).
2. F. Li et al., mAbs. 2 (5) 466-479 (2010).
3. The Merck Index (Merck & Co. Inc., Whitehouse Station, NJ, 13th ed., 2001).
4. P. Hossler et al., Biotechnol. Prog. 29 (4) 1023-1033 (2013).
5. S.S. Bhadoriya et al., Biochem. Pharmacol. 2 (2) 1000113 (2013) doi:10.4172/2167-0501.1000113.
6. K.H. Bhandari et al., Biol. Pharm. Bull. 30 (6) 1171-1176 (2007).
7. J. Sruti et al., Indian J. Pharm. Sci. 75 (1) 67-75 (2013).
8. R. McPherson and A. Gauthier, Biochem. Cell Biol. 82 (1) 201-211 (2004).
9. D.M. Phillips and R.C. Haut, J. Orthop. Res. 22 (5) 1135-1142 (2004).
10. G. Serbest et al., FASEB J. 20 (2) 308-310 (2006).

About the Author
Ping Xu is a scientist at Bristol-Myers Squibb, Process Development, 519 Route 173 West, Bloomsbury, NJ 08804, ping.xu@bms.com; Xiao-Ping Dai is director of Process Development, Biologics at Celgene Corporation, 200 Connell Drive, Berkeley Heights, NJ 07922, and previously manager of Cell Culture Sciences at Bristol-Myers Squibb when this work was done; Albert Kao is an undergraduate student at the University of Michigan, Ann Arbor, MI 48109; Rosario Scott is a senior scientist, and Reb Russell is site head and head of Biologics Process Development, both at Bristol-Myers Squibb, Process Development, 519 Route 173 West, Bloomsbury, NJ 08804.