262
Biotechnol. Prog. 1990, 6, 262-265
Enhancing Oxygen Transfer in Surface-Aerated Bioreactors by Stable Foams Lu-Kwang Ju’ and William B. Armiger BioChem Technology, Inc., 66 Great Valley Parkway, Malvern, Pennsylvania 19355
To enhance oxygen transfer in surface-aeration bioreactors, stabilized foams were generated to increase the gas-liquid interfacial area by slowly introducing coarse bubbles into media containing fetal bovine serum. The bubble sparging rates were so low (i.e., 20 and 50 mL/h) that the contribution to oxygen transfer from these bubbles was due to foaming instead of bubbling. Furthermore, no physical cell damage caused by bubble sparging was observed. Oxygen transfer coefficients, k ~ ain, the bioreadors were measured in cell-free media. Without the foam-stabilizing agent (i-e.,serum), no appreciable change a observed due to the bubble sparging. On the other hand, with serum, k ~ a in k ~ was increased with increasing serum content and bubble sparging rate and corresponded well with the degree of foaming. With 10% fetal bovine serum and a bubble sparging a approximately 90% compared with no foaming. The rate of 50 mL/h, k ~ increased enhancing effect of foam on oxygen transfer in surface aeration bioreactors has been further demonstrated with hybridoma cultures simultaneously grown in three identical bioreactors with and without stabilized foams.
Introduction Oxygen transfer has always been a central topic for design and scale-up of aerobic bioprocesses. It plays an even more important role in cell culture. To reduce the cost of producing mammalian cell products on a large scale, it is desirable to maximize the cell density. With high cell densities, the product concentration increases, which reduces the cost of purifying the product. Furthermore, no additional serum components are needed at higher cell densities, since serum components are thought to catalyze cellular reactions and are not significantly depleted in the medium (I). Therefore, at higher cell densities, the yield of product from serum increases. Since serum is the predominant cost of culture media, high cell densities greatly reduce the cost of producing mammalian cell products. Mammalian cells require oxygen to synthesize ATP via oxidative phosphorylation. They will neither grow nor remain viable for extended periods of time under anaerobic conditions. High cell density can only be achieved when each cell’s oxygen demand is satisfied. In addition, Kilburn et al. (2) have reported that the growth rate, respiration rate, and carbohydrate metabolism are also functions of dissolved oxygen Concentration. In fact, the ability to maintain an adequate oxygen concentration is important in optimizing product yields, since it determines the nutrient conversion efficiency into wanted product. In small-scale cell cultivation systems, oxygen is usually supplied by mass transfer through the gas-liquid interface at the culture surface. Mild agitation, besides aiding this process, also provides the necessary mixing to keep the culture more homogeneous. With mild agitation, the oxygen transfer in surface-aeration bioreactors is controlled by the liquid-side film resistance of the gas-liquid interface at the culture surface. Therefore, the oxygen transfer rate,
OTR, can be expressed as where kL is the liquid-side oxygen transfer coefficient, a is the interfacial area per unit volume of medium, C* is the saturation concentration of dissolved oxygen at the gasliquid interface, and CL is the concentration of dissolved oxygen in the bulk liquid phase. It is widely recognized that surface-aerated bioreactors are strongly limited by the interfacial area, a, which decreases with increasing scale. At large scales, sufficient OTR cannot be obtained without other modifications. High-rate air sparging of microbial fermentations increases the gas-liquid interfacial area. Unfortunately, it is not adaptable to cell culture, since air sparging damages animal cells and foams the medium (3-5). Foaming is generally viewed as undesirable in cell culture systems. However, stabilized foams on the culture surface significantly increase the gas-liquid interfacial area, which, in turn, enhances the oxygen transfer in surface-aerated bioreactors. This work improves the performance of surface-aerated bioreactors by generating well-controlled foams on the culture surface.
Materials and Methods Hybridoma. Hybridoma 14-4-48, ATCC HB32, was used in this study for its well-documented optimal growth environment and specific oxygen uptake rate (6-8). Derived by fusing SP2/0-Ag14 cells with lymphocytes of C3H cells, the cell line produces a cytotoxic monoclonal antibody, IgGzak, that reacts with the I-Ek and Ck determinants. Culture Medium. The base medium used in the study was Dulbecco’s modified Eagle’s medium (DMEM) with glutamine and high glucose (Mediatech, Washington, DC). The medium was further supplemented with the following substances added per liter of DMEM: 2.2 g of sodium
8756-7938/90/3006-0262$02.50/0 0 1990 American Chemical Society and American Institute of Chemical Engineers
263
Biotechnol. Prog., 1990, Vol. 6, No. 4
bicarbonate, 10 mL of nonessential amino acids, 100 000 units of penicillin, 100 mg of streptomycin, 13.2 mg of oxaloacetic acid, 0.8 mg of insulin, 5.5 mg of pyruvic acid, and 10% fetal bovine serum (Hazelton Research Products, Inc., Lenexa, KS). Cells were inoculated a t a density of (1-2) X lo5 viable cells/mL. Bioreactors and Equipment. Hybridoma cultures were grown in three identical 1-L double-side-arm Celstir bioreactors (Wheaton Scientific, Millville, NJ) a t 37 f 0.1 "C. Each bioreactor was equipped with one pH electrode and one dissolved oxygen electrode. To maintain the culture pH value around 7.2, the head space of each bioreactor was purged a t a rate of 100 mL/min with gas of composition 20 9% 0 2 , 5 76 COz, and 75 9% Nz by volume. Assays. Viable and total cell concentrations were assessed by staining samples with Trypan blue (Sigma) and counting presumed dead and live cells in a hemocytometer by using a light microscope. Glucose and lactate concentrations were determined by HPLC (Model 5000, Varian Associates, Inc., Walnut Creek, CA) with an Aminex HPX-87H column (Bio-Rad Laboratories, Richmond, CA). Antibody titers were determined by using an enzymelinked immunosorbent assay (ELISA) following the procedures of Bosworth et a1 (9). Volumetric Oxygen Transfer Coefficient, kLa. Measurements of kLa in bioreactors were made with the dynamic method proposed by Dang e t al. (10) and improved by Ruchti et al. (11). Following a step change of the influent gas from nitrogen to the 5 % COZ gas mixture used in hybridoma cultures, the variation in oxygen partial pressure in the liquid was measured by an oxygen electrode, recorded on computer discs with a data acquisition and analysis software (FERMAC software, BioChem Technology, Inc., Malvern, PA), and analyzed to generate a semilogarithmic plot of Yp vs time. The dimensionless oxygen partial pressure, Yp, is defined as follows:
where P is oxygen partial pressure and the subscripts g, 1, and N represent the gas mixture, liquid, and nitrogen, respectively. Figure 1 shows a typical absorption curve and the corresponding plot of In Yp vs time. Based on a linear model that accounts for the gas residence time and the serial resistances of oxygen transfer through the liquid film and the electrode membrane, the time where Yp = l / e (i.e., tile) is equal to the sum of the time constants (11, 12): (3) where TEis the electrode time constant, TFis the liquidfilm time constant, and TGis the mean gas residence time. Thus kLa can be determined, provided the values of TE, T F ,and TGare known. The value of T E+ TFcan be evaluated experimentally from the electrode response curve to a step change of liquid oxygen partial pressure. The electrode was first placed in an external solution sparged with nitrogen to obtain a steady value of PN. It was then moved quickly into the bioreactor agitated a t 100 rpm and saturated with the gas mixture used in the study. The switching time was approximately 2 s. The value of TE+ TFwas assessed by subtracting the switching time from til, obtained from the electrode response curve. Finally, TG can be estimated by dividing the head-space volume by the volumetric gas flow rate.
loo>
-0.2
-
-0.4 n
-
,-0.6
>
W
C =d
-0.8-1
.o -
-1.2
1
1
I
Figure 1. Assessment of kLn in a bioreactor: (a) typical absorption curve of oxygen in t h e culture medium; (b) corresponding plot of In Y pversus time for determination of t l / e .
Results and Discussion Bioreactor kLa. A foam can be considered as a type of emulsion in which the inner phase is a gas. As with all emulsions, a surfactant is necessary t o give stability. Although a huge variety of surfactants are available as foaming agents, fetal bovine serum (FBS) has been used in this study as the foam stabilizer to avoid possible problems of toxicity and/or of cell metabolism changes from introducing other surfactants. To study the effect of well-controlled foams on oxygen transfer in surface-aerated bioreactors, kLa measurements were made simultaneously in three bioreactors with cellfree media containing 0, 3, and 10% (v/v) fetal bovine serum. The same bioreactor configuration and agitation speed, 100 rpm, were employed to minimize effects from factors other than foaming. While the gas flow rate in the bioreactor head space was kept at 100 mL/min, the coarsebubble sparging rates for generating foams were 0,20, and 50 mL/h in different runs. As expected, no stable foams were observed in the serum-free system at all sparging rates. On the other hand, 1-2 foam layers were found in systems containing FBS with coarse-bubblesparging. Foam usually accumulated selectively on one side of the bioreactor due to influences of electrodes, sampling lines, etc. Experimental results are summarized in Figure 2. kLa did not change appreciably with FBS content when no bubble sparging was employed. Furthermore, in the serumfree system, where the sparged bubbles were not stabilized, the air sparging had only a minute effect on kLa. This can be explained by the following analysis for a stagnant system. The bubble diameter was about 0.5 cm, with a rise velocity through stagnant water reported to be about 25 cm/s (13). Without being stabilized a t the liquid surface, each bubble contacted the medium for only 0.32
264 4.5
,
2.0
I
Biotechnol. Prog., 1990,Vol. 6, No. 4
,
1
0
2.0 Y;
0
i I
I
I
I
I
2 4 6 8 Fetal Bovine Serum Content
'
(w)
1'0
Y
1.0
~
u I
7
Control I
A
I
6
F e t a i Bovine ' Serum -Conten:
I
8
(w)
in ,
Y
Figure 2. Experimental results of kLa as a function of the fetal bovine serum content of the culture media in the surfaceaerated bioreactors sparged with coarse bubbles at 0 (control), 20, and 50 mL/h.
Figure 3. k ~ ratios a (serum containing:serumfree) as a function of the fetal bovine serum content of the culture media in the surface-aerated bioreactors subjected t o different coarsebubble sparging rates.
s since the liquid depth for bubble travel was about 8 cm. With a gas sparging rate of 50 mL/h, the time-averaged gas-liquid interfacial area available for oxygen transfer a t any instant due to the bubbling can be estimated as
within *0.05% of the indicated range. Throughout the study, no apparent cell damage was observed in cultures subjected to the low-frequency, coarse bubble sparging. Figure 4 shows the dissolved oxygen profile and livecell density obtained in hybridoma cultures sparged with coarse bubbles a t 0,20, and 50 mL/h. The bioreactor with surface aeration alone became oxygen-limited, since the dissolved oxygen levels quickly dropped to zero. In the bubble-sparged cultures, oxygen limitation was delayed about 10 h. Qualitatively, this indicates that higher oxygen transfer rates were achieved with coarse-bubble sparging. To assess the enhanced oxygen transfer rate more quantitatively, the live-cell densities when the dissolved oxygen contents just reached zero were 1.05 X lo6, 1.87 X lo6, and 1.95 X lo6 cells/mL with bubble sparging rates of 0, 20, and 50 mL/h, respectively. Assuming the same specific oxygen demand for cells with and without coarsebubble sparging and that the pseudo-steady-state oxygen transfer rate equals the oxygen consumption rate, the oxygen transfer enhancement factors can be estimated as the ratios of these viable cell concentrations, Le., 1.78 and 1.86 with bubble sparging rates of 20 and 50 mL/h, respectively. These results are very consistent with the kLa ratios (see Figure 3) obtained in the cell-free systems, i.e., 1.86 and 1.89 for bioreactors with 20 and 50 mL/h bubble sparging, respectively. The changes in glucose, lactate, and monoclonal antibody concentration during the cell cultivation are shown in Table I. The monoclonal antibody production in bubblesparged systems was about 40% higher than the control up to about 50 h. However, judging from the dissolved oxygen profiles, cells in the control bioreactor appeared to survive better and longer under the oxygen-limited condition. This is further supported by the observation t h a t the rates of glucose consumption and lactate production a t the final cultivation stage were higher in the control system. About 2 mol of lactate was produced/ mol of glucose consumed, just as expected from normal glycolysis under anaerobic conditions. On the other hand, cells in the bioreactors with stable foams started to lose their metabolic activities after being subjected to anaerobic conditions for a shorter period of time. Consequently, the increase in the final product formation in systems with bubble sparging was insignificant. The shorter survival period of cells under anaerobic conditions in bioreactors with stable foams could result from reaching certain limitations imposed by the higher cell densities present in these systems (e.g., depletion of
= 0.053 cm2 When compared with the medium surface area, A , = 79 cm2 (the bioreactor diameter is about 10 cm), it represents less than 0.1 % increase in interfacial area. Although the actual situation is more complicated since the mild agitation slightly increases the bubble retention time in the medium, it is apparent that, without being stabilized, bubbles did not contribute much to the overall gasliquid interfacial area. Consequently, the significant k ~ a increase when coarse bubbles were sparged through FBScontaining media can only be attributed to the larger gasliquid interfacial area in stable foams. Figure 3 shows the ratios of k ~ between a the serumcontaining and the serum-free systems plotted against the serum content. This more clearly illustrates the effects of serum content and bubble sparging rate on the degree of foaming, which in turn determines the interfacial area and oxygen transfer rate. The degree of foaming increases with increasing bubble sparging rate and the serum content. A kLa ratio as high as about 1.9 was achieved in the bioreactor with 10% fetal bovine serum when sparged a t the rate of 50 mL/h. The feasibility of improving oxygen transfer in surface-aerated bioreactors by providing stabilized foams to increase the interfacial area has been demonstrated. Hybridoma Culture Study. To demonstrate stable foams are useful in real cultivation processes, hybridoma cultures were grown simultaneously in three identical bioreactors. The same bioreactor configuration and agitation speed, 100 rpm, were employed. The only difference among them was the foam-generating mechanism. No foaming was provided in the control culture, while in the other two bioreactors 1-2 foam layers were maintained on the culture surfaces by gently sparging coarse gas bubbles a t the rates of 20 and 50 mL/h. The bubble diameter was about 0.5 cm. It has been reported that large bubbles usually cause much less physical damage to shearsensitive cells than small bubbles (5, 14). The low gas delivery rates were accomplished by a dual-chamber peristaltic pump (Model 420, BioChem Technology, Inc., Malvern, PA) that has a wide flow-rate range from 1.0 to 999.9 mL/h in 0.1 mL/h increments with an accuracy
265
Biotechnol. Prog., 1990, Vol. 6, No. 4 A
Figure 4. Typical dissolved oxygen time profiles and live-cell density in hybridoma cultures with different coarse-bubble sparging rates: X, 0 mL/h (control); A, 20 mL/h; 0 , 50 mL/h. Table I. Changes in Glucose, Lactate, and Monoclonal Antibody Concentrations i n Hybridoma Cultures. glucose concn, lactate concn, monoclonal antibody time, g/L g/L titer, mg/L h A B C A B C A B C 0
38.5 49.5 66.0
3.31 2.31 1.70 0.84
3.27 2.39 1.71 1.14
3.27 2.50 1.74 1.25
0.82 1.76 2.18 3.10
0.86 1.72 2.23 2.44
0.86 5.19 1.50 7.08 2.13 9.08 2.31 16.24
8.01 11.64 13.45 18.04
4.84 7.69 10.25 17.21
Hybridoma cultures were grown simultaneously in three identical bioreactors. In A, no stable foams were present, while in B and C, stable foams were generated by gently sparging coarse bubbles at 20 and 50 mL/h, respectively.
essential nutrients and accumulation of toxic wastes). Better control of other process conditions like pH, temperature, and agitation might improve product yields, since our experience shows that cells exposed to anaerobic conditions in final batch-culture stages are much more sensitive to fluctuations in the cultivation environment. Nevertheless, the increased oxygen transfer capability of surface-aerated bioreactors with stable foam layers on the culture surface has been clearly demonstrated.
Literature Cited (1)Glacken, M. W.; Fleischaker, R. J.; Sinskey, A. J. Largescale Production of Mammalian Cells and their Products: Engineering, Principles and Barriers to Scale-up. Ann. N.Y. Acad. Sci. 1983, 413, 355. (2) Kilburn, D. G.; Lilly, M. D.; Self, D. A.; Webb, F. C. J. Cell Sci. 1969, 4, 25. (3) Kilburn, D. G.; Webb, F. C. The Cultivation of Animal Cells at Controlled Dissolved Oxygen Partial Pressure. Biotechnol. Bioeng. 1968, 10, 801. (4) Telling, R. C. Submerged Culture of Hamster Kidney Cells in a Stainless Steel Vessel. Biotechnol. Bioeng. 1965, 7, 417. (5) Handa, A.; Emery, A. N.; Spier, R. E. On the Evaluation of Gas-Liquid Interfacial Effects on Hybridoma Viability in Bubble Column Bioreactors. Deu. Biol. Stand. 1987,66,241. (6) MacMichael, G.; W. B.; Armiger, J. F.; Lee, R. Mutharasan, On-Line Measurement of Hybridoma Growth by Culture Fluorescence. Biotechnol. Tech. 1987, I , 213. (7) Ozato, K.; Mayer, N.; Sachs, D. H. Hybridoma Cell Lines Secreting Monoclonal Antibodies to Mouse H-2 and I, Antigens. J. Immunol. 1980,124, 533. (8) MacMichael, G. The Use of Continuous Culture to Optimize the Hybridoma Culture Environment. Fed. Proc. 1987,46,530. (9) Bosworth, J. M., Jr.; Brimfield, A.; Naylor, J. A.; Hunter, K. W., Jr. Measurement of Monoclonal Antibody Concentrations in Hybridoma Cultures: Comparison of Competitive Inhibition and Antigen Capture Enzyme Immunoassays. J. Immunol Methods 1983, 62, 331. (10) Dang, N. D. P.; Karrer, D. A.; Dunn, I. J. Oxygen Transfer Coefficients by Dynamic Model Moment Analysis. Biotechnol. Bioeng. 1977, 19, 853. (11) Ruchti, G.; Dunn, I. J.; Bourne, J. R. Comparison of Dynamic Oxygen Electrode Methods for the Measurement of k ~ a .Biotechnol. Bioeng. 1981,23, 277. (12) Sobotka, M.; Prokop, A.; Dunn, I. J.; Einsele, A. Review of Methods for the Measurement of Oxygen Transfer in Microbial Systems. Annu. Rep. Ferment. Processes, 1982; 5, 127. (13) Heijnen, J. J.; Van’t Riet, K. Mass Transfer, Mixing and Heat Transfer Phenomena in Low Viscosity Bubble Column Reactors. Chen. Eng. J . 1984, 28, B21. (14) Lavery, M.; Kearns, M. J.; Price, D. G.; Emery, A. N.; Jefferis, R.; Nienow, A. W. Physical Conditions During Batch Culture of Hybridomas in Laboratory Scale Stirred Tank Reactors. Deu. Biol. Stand. 1985, 60, 199.
Accepted May 29,1990. Registry No. 02, 7782-44-7.
