Transcriptomic Responses to Sodium Chloride-Induced Osmotic Stress: A Study of Industrial Fed-Batch CHO Cell Cultures Duan Shen Dept. of Chemical and Biological Engineering and Center for Biotechnology and Interdisciplinary Studies, Rensselaer Polytechnic Institute, Troy, NY 12180
Thomas R. Kiehl Multidisciplinary Science Program and Center for Biotechnology and Interdisciplinary Science, Rensselaer Polytechnic Institute, Troy, NY 12180
Sarwat F. Khattak and Zheng Jian Li Process and Product Development, Technical Operations, Bristol-Myers Squibb Company, Syracuse, NY 13057
Aiqing He, Paul S. Kayne, Vishal Patel, and Isaac M. Neuhaus Applied Genomics Dept., Bristol-Myers Squibb Company, Pennington, NJ 08534
Susan T. Sharfstein Dept. of Chemical and Biological Engineering and Center for Biotechnology and Interdisciplinary Studies, Rensselaer Polytechnic Institute, Troy, NY 12180 DOI 10.1002/btpr.398 Published online February 8, 2010 in Wiley Online Library (wileyonlinelibrary.com).
The rapidly expanding market for monoclonal antibody and Fc-fusion-protein therapeutics has increased interest in improving the productivity of mammalian cell lines, both to alleviate capacity limitations and control the cost of goods. In this study, we evaluated the responses of an industrial CHO cell line producing an Fc-fusion-protein to hyperosmotic stress, a well-known productivity enhancer, and compared them with our previous studies of murine hybridomas (Shen and Sharfstein, Biotechnol Bioeng. 2006;93:132–145). In batch culture studies, cells showed substantially increased specific productivity in response to increased osmolarity as well as significant metabolic changes. However, the final titer showed no substantial increase due to the decrease in viable cell density. In fed batch cultures, hyperosmolarity slightly repressed the cellular growth rate, but no significant change in productivity or final titer was detected. To understand the transcriptional responses to increased osmolarity and relate changes in gene expression to increased productivity and repressed growth, proprietary CHO microarrays were used to monitor the transcription profile changes in response to osmotic stress. A set of osmotically regulated genes was generated and classified by extracting their annotations and functionalities from online databases. The gene list was compared with results previously obtained from similar studies of murinehybridoma cells. The overall transcriptomic responses of the two cell lines were rather different, although many functional groups were commonly perturbed between them. Building on this study, we anticipate that further analysis will establish connections between productivity and the expression of specific gene(s), thus allowing rational engineering of mammaC 2010 American Institute of lian cells for higher recombinant-protein productivity. V Chemical Engineers Biotechnol. Prog., 26: 1104–1115, 2010 Keywords: hybridoma, Chinese hamster ovary cell, microarray, transcriptome, osmotic stress
Introduction Monoclonal antibodies (mAbs) have been widely applied as diagnostic and therapeutic reagents for more than two Additional Supporting Information may be found in the online version of this article. Correspondence concerning this article should be addressed to S. T. Sharfstein at this current address: College of Nanoscale Science and Engineering, University at Albany, Albany, NY 12203; e-mail:
[email protected]. 1104
decades. The rapidly increasing demand for monoclonal antibodies, stimulated by market expansion for the existing products and by the continuous launch of new products and new applications, has placed significant pressure on process development and manufacturing to improve productivity. Substantial improvements in viable cell density and culture longevity have been accomplished in the last 20þ years. As reviewed by Wurm, in 1986, cultured mammalian-cell bioreactors typically reached a peak cell density of 2 106 cells/mL and cultures were maintained for approximately C 2010 American Institute of Chemical Engineers V
Biotechnol. Prog., 2010, Vol. 26, No. 4
7 days.1 In contrast, in 2004, typical final cell densities were 10 106, and cultures were maintained for almost 3 weeks with a high level of viability, yielding a 17-fold increase in the integrated viable cell number. One of the most significant changes in bioprocessing has been the transition from batch to fed-batch processes in which concentrated nutrient feeds are supplemented throughout the culture period.2 This permits a steady supply of nutrients while reducing waste product accumulation that may be inhibitory both to cell growth and antibody productivity. Despite the continuing supply of nutrients, fed-batch cultures eventually exhibit decreases in viability. Although there may be a variety of reasons for this decrease in viability, including accumulation of lactate and ammonia, the increase in osmolarity due to the concentrated nutrient feeding, the production of lactate, and the addition of base to maintain pH has been hypothesized to play a significant role.3,4 A variety of process strategies have been explored to improve specific productivity from mammalian cell cultures including the addition of a variety of compounds such as mouse peritoneal factors,5 growth inhibitors,6 autocrine factors,7 and cyclic mononucleotides.8 Application of environmental stresses such as alterations in temperature9,10 and dissolved oxygen (DO) tension11 have also been employed. Two of the most widely studied approaches to increase specific productivity are application of hyperosmotic stress and addition of sodium butyrate. Hyperosmotic stress can be easily induced by addition of salts or sugars to the culture medium, and its effect on increasing the specific productivity has been observed in many hybridoma cell liness12–14 as well as several Chinese hamster ovary (CHO) cell lines.15,16 However, the increase in specific productivity does not, in general, result in a substantial increase in overall yield due to depressed cell growth and the resulting decrease in viable cell number, and the extent that antibody productivity is increased is cell-line specific.17,18 A critical question is whether the mechanisms by which osmotic stress increases specific antibody production are similar in different hybridoma cell lines and between hybridoma and CHO cell lines. To understand the mechanisms for the increase in antibody production in response to hyperosmotic stress, the responses of many cellular processes including cellular metabolism,19,20 RNA transcription,15,21 protein translation,22 post-translational processing,22 and protein secretion21 have been examined. Recently, Lee et al. reported a proteomic analysis of CHO cells in response to hyperosmotic pressure.16 In this study, they found three identifiable proteins that were differentially expressed: tubulin, which was downregulated, pyruvate kinase and glyceraldehyde-3-phosphate dehydrogenase, which were upregulated. However, to date, little information on the global transcriptional response has been obtained. Investigations of the response of mammalian cell culture to culture conditions have been greatly facilitated by the rapid advance of genomic and proteomic gene-expression profiling techniques. DNA microarrays and proteomic analysis have been used to investigate global gene expression of mammalian cells under various conditions including mouse hybridoma MAK under metabolic shift,23 MAK and CHO cells under sodium butyrate treatment24 and temperature shift,25 and NS0 cells at reduced temperature.26 These studies provided some information about changes at the protein level, but the transcriptomic information generated was substantially greater due to higher sensitivity, better coverage of DNA microarrays than proteomic techniques, and more in-
1105
formation available in genomic databases than proteomic databases. Using proprietary CHO arrays, we probed the geneexpression profiles of CHO cells producing an Fc-fusion protein under sodium chloride-induced hyperosmotic stress in fedbatch cultures. Sodium chloride was selected as the osmolyte for several reasons. First, it has been widely shown that sodium chloride-induced osmotic stress can lead to increased specific productivity. In addition, most fed-batch cultures require pH control by the addition of base (generally sodium hydroxide or sodium bicarbonate) late in culture; hence, sodium increases are typical in fed-batch cultures. Finally, our previous study of hybridoma responses to increased osmolarity also employed sodium chloride, allowing better comparisons of the results from this study with our previous study of the global transcriptional responses of murine hybridoma OKT327 than if a different osmolyte had been applied. We also compared the differentially expressed genes identified in this study with differentially expressed genes reported in response to other culture-environment perturbations (e.g., butyrate, temperature shifts). Expression profiles of selected genes generated by microarray analysis were validated using TaqManV realtime quantitative PCR. Differentially expressed genes were categorized into several functional groups, e.g., transcriptional regulation, cell-cycle regulation, etc., to obtain a better picture of the physiological changes that occur in response to hyperosmotic stress. Understanding these physiological changes is the first step in the effective development of cellular and molecular strategies for increasing antibody production by mammalian cells. The results presented here will be used to better understand the rate-limiting processes in the production of monoclonal antibodies and Fc-fusion proteins and to optimize cell lines and process conditions to improve productivity. R
Materials and Methods Cell lines and culture condition Proprietary Chinese hamster ovary cell line B0, which produces a recombinant Fc-fusion protein, was kindly provided by Bristol-Myers Squibb (BMS, Syracuse, NY). B0 was cultured in a proprietary animal-derived-component-free (ADCF) medium. The B0 cell culture was performed at the BMS facility in Syracuse, NY. For batch culture, 150 mL of B0 cells were cultured in 500 mL spinner flasks (Corning, Corning, NY) at 37 C and 150 rpm in a humidified 5% CO2 incubator. Cells were routinely subcultured from late exponential phase cultures and inoculated at 4 105 cells/mL. Samples were taken on a daily basis. The integrated viable cell density R t (IVCD) over the culture period is defined as IVCD ¼ 0 xdt, where x is the viable cell density, which varies over the culture period from inoculation to culture time t. In our study, this integration was approximated by the following trapezoidal formula:
IVCD
X1 2
ðxi þ xiþ1 Þðtiþ1 ti Þ:
Fed-batch cultures of B0 were performed in a 5-L bioreactor with a 1.5-L working volume of the same medium. B0 cells expanded in batch culture were inoculated at an initial seeding density of 1.5 105 viable cells/mL. A proprietary, enriched feeding medium was used during the fed-batch process. Temperature, pH, and DO were monitored and recorded throughout the fed-batch cultures. Agitation speed was set at 150 rpm. Gas flows were controlled via
1106
Biotechnol. Prog., 2010, Vol. 26, No. 4
rotameters. Air and CO2 were sparged into the bioreactor through a submerged frit. Oxygen was sparged through the same frit for DO control. Air was also supplied through the head space for CO2 removal. pH control was performed by sparging with CO2 or addition of 1 N NaOH. Samples were taken on a daily basis for both batch and fed-batch cultures of B0. Cell counts and cell viability determination were performed using a Cedex AS20 cell analyzer (Innovatis, Bielefeld, Germany). Glucose, glutamine, lactate, and ammonium concentrations were analyzed using a Bioprofile 200 cell culture analyzer (NOVA Biomedical, Waltham, MA).
Osmotic stress For batch cultures, hyperosmotic stress was applied either at 24 h after inoculation (early onset) or 70 h after inoculation (late onset) at varying osmolarities. Based upon the results from the batch studies (described below), hyperosmotic stress was induced at approximately 70 h after inoculation for the fed-batch cultures. Sterile 5 M NaCl was added into the medium to raise the osmolarity to the set level (the target increase was 100 mOsm for fed-batch cultures; however, due to experimental variation, measured increases ranged from 70 to 100 mOsm). The elevated osmolarity was maintained for the duration of the culture. The control cultures were maintained in the standard medium (osmolarity ca. 290 mOsm). The osmolarity was estimated by NOVA analyses and verified using a VAPRO5520 Vapor pressure osmometer for selected time points (Wescor, Logan, UT).
Product titer assay Samples were centrifuged at 1,000 rpm for 5 min and the supernatants were stored at 20 C before product titer assay. Fc-fusion protein concentrations of B0 culture were analyzed using the Easy-TiterV human IgG (gamma chain) assay kit (Pierce, Rockford, IL), according to the manufacturer’s instructions. The assay protocol has been described previously.27 Standards and samples were assayed in triplicate. Product titers are reported as a percentage of the isotonic control from the same experiment; typical fed-batch titers for this molecule are in the range commonly seen for industrially produced antibodies and fusion proteins. R
lyzer with the RNA6000 LabChipV kit as per manufacturer’s instructions. R
Microarray processing Custom-made AffymetrixV CHO microarrays were used for the transcriptome analysis of B0. The construction of CHO cDNA libraries and microarray was done by the Consortium for CHO Genomics and has been previously described.28 The array processing and scanning were performed by the Applied Genomics Department of BristolMyers Squibb Company (Pennington, NJ). In brief, double-stranded cDNAs were synthesized from total RNA by using the cDNA Synthesis System (Invitrogen, Carlsbad, CA). Biotin-labeled cRNA was generated and used to probe the microarrays according to the manufacturer’s recommendations (Affymetrix). Orthologous genes in the murine MOE430A and CHO arrays were identified using the Basic Local Alignment Search Tool (BLAST29). The sequences of differentially expressed MOE430A probes were compared with all the sequences on the CHO microarray, whereas the sequences of differentially expressed CHO probes were compared with the mouse RefSeq database. R
Data acquisition and processing Acquisition of the quantified signal intensities from the array images was performed using the Affymetrix Microarray Suite (MAS 5.0). Chip signals of OKT3 samples were normalized using the GC-RMA algorithm30 within the GeneTraffic software package (Iobion Informatics, La Jolla, CA). The chip signals of B0 samples were normalized using the RMA algorithm30 with the BMS in-house software package. In both cases, signals of the control replicates at each time points were used as the baseline. To filter out the significantly regulated genes, all probe sets with a signal lower than 50 were eliminated from further analysis. One-way ANOVA was performed and genes with P-value lower than 0.05 were considered as statistically different. The normalized chip signals of the subset with statistically significant changes were filtered according to the fold change at each time point. Genes with twofold or greater change at any time point in any reactor set were selected, and the subset of genes with an average change of 1.5-fold or greater was considered differentially expressed. Quantitative RT-PCR
RNA isolation Samples containing 1 107 cells were taken from each of the control or stressed fed-batch cultures at 1, 2, 4, 8, 24, 48 h after the addition of NaCl solution (i.e., at 71, 72, 74, 78, 94, and 118 h in culture). The time points were selected based upon our previous observations that approximately 7 to 12 h was required for the increase in antibody synthesis in response to hyperosmotic stress.22 Ten million cells from each sample were taken for total RNA purification using the RNeasyV Mini kit (Qiagen, Valencia, CA) as per manufacturer’s instructions. The concentration of RNA was determined by measuring the A260 of the RNA solution in 10 mM Tris-EDTA buffer (pH 8.0). RNA purity was verified by measuring the A260/A280 ratio. Values for all RNA samples were between 1.8 and 2.0. RNA integrity was further verified using an Agilent 2100 BioanaR
The microarray results for selected genes were validated using TaqManV 50 -nuclease real-time quantitative RT-PCR assays. Custom TaqManV Gene Expression Assays for selected genes were obtained from Applied Biosystems. Total RNA, purified as described above, was diluted and then directly used in one-step quantitative RT-PCR. The real-time PCR reaction was performed in an ABI 7900HT real-time PCR system (Applied Biosystems, Foster City, CA) using TaqManV One-Step RT-PCR Master Mix Reagents Kit as per manufacturer’s instruction. Reactions were performed in duplicate or triplicate for each biological replicate sample at each time point. Results from replicate reactions were averaged and used as the value for each biological replicate. Variations between replicate reactions were small in comparison with variations between biological replicates and were neglected in error analysis. RNase free water R
R
R
Biotechnol. Prog., 2010, Vol. 26, No. 4
1107
was added to the PCR mixture instead of cDNA as a negative control. To obtain the relative gene expression level in each sample, a standard qPCR curve for each tested gene was constructed using serial dilutions of selected total RNA sample. Statistical analysis of microarray and qPCR data was performed using GraphPad Prism version 4.03 for Windows, GraphPad Software, San Diego, California, www.graphpad.com.
Results and Discussion Cell growth, productivity, and metabolic changes in batch cultures To determine the optimal increase in osmolarity and the optimal time for induction of osmotic stress, batch cultures were performed with varying increases in osmolarity (0–160 mOsm above control) and an early (24 h after inoculation) or late (70 h after inoculation) induction of osmotic stress. As shown in Figure 1A and Table 1 for the 24-h induction, osmotic stress decreased the specific growth rate and maximum viable cell density in a dose-dependent manner. The specific productivities were similarly increased in a dose-dependent manner (Figure 1B); however, no significant increase in titer was observed as the effects of increased productivity were offset by the decrease in IVCD. The metabolic responses to increased osmolarity are shown in Table 2. The specific glucose and glutamine consumption rates were increased in an approximately dose-dependent manner; whereas the yield of lactate from glucose was essentially unchanged in the hyperosmotic cultures. An additional batch culture test was performed with later initiation of osmotic stress at 70 h after inoculation, as shown in Figure 2. Not surprisingly, the increase in specific productivity was less substantial. The specific productivities of the cultures with 100 and 145 mOsm increases in osmolarity were 152% and 219% of the isotonic culture, respectively. However, the later onset of osmotic stress showed a less substantial reduction of cell growth. The maximum viable cell densities of cultures with 100 and 145 mOsm increases in osmolarity were 89% and 71% of control cultures, respectively. Most significantly, the final titers of the cultures increased by 100 and 145 mOsm were 119% and 125% of control cultures, respectively, indicating the overall effect of hyperosmotic stress on the final titer of culture was more significant upon later initiation of stress. Consequently, a 70-h postinoculation time was selected for induction of osmotic stress in the fed-batch bioreactors. As fed-batch cultures typically show an increase in osmolarity due to feeding (see, for example, Figure 3C), an increase of 100 mOsm was selected rather than the 145 mOsm for the fed-batch studies. Cell growth, productivity, and metabolic changes in fed-batch bioreactors The production of monoclonal antibodies and Fc-fusion proteins in the biopharmaceutical industry is routinely performed in fed-batch cultures. Hence, we selected fed-batch bioreactor cultures for our transcriptional analysis of CHO cell responses to increased osmolarity. Figure 3 shows the typical performance comparison of B0 cells in a standard fed-batch process with and without osmotic shock. As we observed in batch culture, hyperosmolarity repressed cell growth. The growth rate and peak viable cell density of stressed cells decreased to 89% and 86% of isotonic culture
Figure 1. Cell growth and Fc-fusion protein production of CHO cell line B0 with osmotic stress applied at 24 h after inoculation. Osmotic shock was applied at 24 h after inoculation by addition of 5 M NaCl solution to increase the osmolarity. A: Viable cell density as a function of culture time. B: Secreted antibody concentration as a function of the time integral of viable cells. Titer has been normalized with the final titer of the isotonic culture as 100%. Slope indicates the specific antibody production rate (lg/cell/h). Data are the average of biological triplicates; error bars show standard deviation. (C) Viability as a function of culture time. Isotonic control (n) and stress of 40 mOsm (~); 80 mOsm (!); 120 mOsm (^); 160 mOsm (*); 200 mOsm (^).
(respectively), yielding an IVCD of 85% of the IVCD of control cultures. However, no significant increase in specific productivity or final product titer was observed, although no decrease was observed in the final titer, either.
1108
Biotechnol. Prog., 2010, Vol. 26, No. 4
Table 1. Comparison of Cell Growth and Antibody Productivity of B0 in Batch Culture with Osmolarity Increased at 24 h After Inoculation Stress Goal 0 (control) 40 mOsm 80 mOsm 120 mOsm 160 mOsm
Growth Rate l (%)*
Osmolarity (mOsm) 335 376 403 443 489
1 10 5 9 9
100 90 84 80 56
Max. Viable Cell Density (%)*
4 2 4 13 5
100 77 59 36 18
9 4 4 3 2
Productivity qp (%)* 100 124 158 239 302
Final Titer (%)*
7 3 7 18 13
100 103 110 105 74
9 0 5 5 0
* All values are given as a percentage of the iso-osmotic control. Values are the average of biological triplicates standard deviation. Table 2. Metabolic Responses of B0 Batch Cultures to Osmotic Stress with Osmolarity Increased at 24 h After Inoculation Stress Goal 0 (control) 40 mOsm 80 mOsm 120 mOsm 160 mOsm
Osmolarity (mOsm) 335 376 403 443 489
1 10 5 9 9
qGluc* 100 120 120 137 136
11 1 12 12 12
qGln* 100 118 112 140 146
12 33 21 52 25
qLac* 100 114 122 154 149
5 2 8 5 11
Ylac/gluc* 100 94 102 112 109
8 1 6 6 7
* All values are presented relative to the iso-osmotic control. Values are the average of biological triplicates standard deviation.
to the batch studies that showed only a small increase in lactate yield, a significant increase in lactate yield was observed in the hyperosmotic fed-batch cultures, particularly, at late culture times. Although the temporal changes in the metabolic patterns varied between different culture replicates, increases in glucose and glutamine consumption under hyperosmotic conditions were seen in all cultures as well as increased yields of lactate from glucose.
Transcriptional changes upon application of hyperosmotic stress
Figure 2. Cell growth and Fc-fusion protein production of CHO cell line B0 with osmotic stress applied at 70 h after inoculation. Osmotic shock was applied at 70 h after inoculation by addition of 5 M NaCl. A: Viable cell density (open symbols) and viability (closed symbols) as a function of culture time. B: Secreted antibody concentration as a function of the time integral of viable cells. Titer has been normalized using the final titer of the isotonic cultures as 100%. Slope indicates the normalized specific antibody production rate (lg/cell/h). Data are average of biological triplicates; error bars show standard deviation. 290 mOsm (isotonic control, n h); 390 mOsm (~ ~); 435 mOsm (l *).
As shown in Table 3, metabolic activity was increased in fed-batch cultures of B0 in a similar manner as seen in batch studies. The uptake of both glucose and glutamine was increased in the hyperosmotic cultures; however, in contrast
The transcriptional profiles of CHO B0 cells grown at isoand hyper-osmolarity were probed using proprietary CHO microarrays. A 1.5-fold change cutoff and a P-value of \0.05 was selected for CHO genes identified as differentially expressed. One hundred and eighty-six probe sets were found differentially expressed; 126 genes were identified from 132 probe sets; the remaining probe sets were for uncharacterized sequences. Table 4 summarizes the number of probe sets and genes found to be differentially expressed in the B0 cell line and compares them with our previous results for murine hybridoma OKT3.27 Detailed expression profiles for these differentially expressed genes can be found in Supporting Information Table 1 (B0) and Supporting Information Table 2 (OKT3). The smaller list of CHO osmotically regulated genes is partly due to the bigger variance between CHO replicate samples; in addition, the CHO arrays contain significantly fewer probe sets (and hence, fewer genes) than the mouse arrays so many differentially expressed genes may not have been detected in these experiments. A hierarchical clustering of the differentially expressed genes is shown in Figure 4. As we observed in our analysis of murine hybridoma responses to increased osmolarity, there are up- and downregulated genes in every gene category with varying temporal patterns as well. The different categories and significantly regulated genes are discussed below. TaqManV qPCR Validation. To verify our microarray results, we used quantitative real-time PCR (qPCR) to analyze five genes from the differentially expressed gene lists. The same RNA samples used for microarray analysis were applied in PCR quantification. The five CHO genes selected for qPCR validation of microarray data were Malt1, an apoptotic gene; Slpi, a protease inhibitor; Slc4a1, a solute transporter; Bcl6, a transcription repressor, and Agt, a salt-stress R
Biotechnol. Prog., 2010, Vol. 26, No. 4
1109 Table 3. Metabolic Responses of B0 Fed-Batch Cultures to Osmotic Stress Culture Time (h)*
qGluc†
qGln†
qLac†
Ylac/Ygluc†
72 97 121 145
111 131 137 173
131 106 142 136
140 180 168 386
107 171 118 284
* Osmotic stress was induced at 70 h after inoculation. are presented relative to the iso-osmotic control.
†
All values
Table 4. Number of Probe Sets and Genes Differentially Expressed Under Hyperosmotic Stress in CHO and Mouse Microarrays with P-value