Biotechnol. Prog. 1990, 6,391-397
39 1
Sparged Animal Cell Bioreactors: Mechanism of Cell Damage and Pluronic F-68 Protection David W. Murhammerf**and Charles F. Goochee*y§ Department of Chemical Engineering, University of Houston, Houston, Texas 77204-4792, and Department of Chemical Engineering, Stanford University, Stanford, California 94305-5025
Pluronic F-68 is a widely used protective agent in sparged animal cell bioreactors. In this study, the attachment-independent Spodoptera frugiperda Sf9 insect cell line was used to explore the mechanism of this protective effect and the nature of cell damage in sparged bioreactors. First, bubble incorporation via cavitation or vortexing was induced by increasing the agitation rate in a surface-aerated bioreactor; insect cells were rapidly killed under these conditions of the absence of polyols. Supplementing the medium with 0.2% (w/v) Pluronic F-68, however, fully protected the cells. Next, cell growth was compared in two airlift bioreactors with similar geometry but different sparger design; one of these bioreactors consisted of a thin membrane distributor, while the other consisted of a porous stainless steel distributor. The flow rates and bubble sizes were comparable in the two bioreactors. Supplementing the medium with 0.2% (w/v) Pluronic F-68 provided full protection to cells growing in the bioreactor with the membrane distributor but provided essentially no protection in the bioreactor with the stainless steel distributor. These results strongly suggest that cell damage can occur in the vicinity of the gas distributor. In addition, these results demonstrate that bubble size and gas flow rate are not the only important considerations of cell damage in sparged bioreactors. A model of cell death in sparged bioreactors is presented.
Introduction Sparging is the most practical method of supplying oxygen to the culture medium in large-scale animal cell bioreactors. Unfortunately, sparging can damage mammalian cells (Kilburn and Webb, 1968; Handa et al., 1987) and insect cells (Murhammer and Goochee, 1988). The specific mechanism of sparging damage is unknown, although it appears to involve bursting bubbles at the medium surface (Handa et al., 1988; Handa-Corrigan et al., 1989) or events in the region of the sparger (Tramper et al., 1987). Addition of the nonionic block copolymer Pluronic F-68 to the culture medium has been shown to protect mammalian cells (Kilburn and Webb, 1968;Handa et al., 1987; Handa-Corrigan et al., 1989; Passini and Goochee, 1989; Reuveny et al., 1986; Radlett et al., 1971) and insect cells (Murhammer and Goochee, 1988, 1990; Maiorella et al., 1988) from the detrimental effects of sparging. Pluronic F-68, which has a total average molecular weight of 8400, consists of a center block of poly(oxypropy1ene) (20?6 by weight) and blocks of poly(oxyethy1ene)at both ends (Stanton, 1957; Swim and Parker, 1960; Schmolka, 1972,1977). The specific mechanism of Pluronic F-68 protection is unclear. Handa et al. (1987,1989) hypothesized that the protection is due to the formation of a stable foam layer on the medium surface into which the cells do not penetrate; thus, the cells are not subject to damage caused by bursting bubbles at the medium surface. We have demonstrated, however, that the presence of a stable foam layer on the medium surface is neither necessary (Murhammer and Goochee, 1988) nor sufficient (Murhammer and Goochee, 1990) for cell protection in a sparged bioreactor.
* Corresponding author. t
University of Houston.
* Current address: Department of Chemical and Biochemical
Engineering, The University of Iowa, Iowa City, IA 52242. 5 Stanford University. 8756-7938/90/3006-0391$02.50/0
We have found that many Pluronic and reverse Pluronic polyols other than Pluronic F-68 can protect insect cells in sparged bioreactors (Murhammer and Goochee, 1990). I t was demonstrated that the ability of a Pluronic or reverse Pluronic poly01 to serve as a protective agent correlateswell with the hydrophilic-lipophilic balance (HLB). HLB is an empirical measure of the emulsifying ability of a surfactant molecule (Griffin, 1949,1954,1955, 1956), and is related to its relative solubility in oil and water. In addition to damage caused by sparging, agitation may also have detrimental effects in animal cell bioreactors. Croughan et al. (1987, 1989) and Cherry and Papoutsakis (1986,1988)have demonstrated that damage to cells on microcarriers results from power dissipation in the form of turbulent eddies. If a relatively large eddy forms in the region of a microcarrier, then the microcarrier is carried along with the flow of the eddy without causing significant cell damage. If the size of the eddy, however, is approximately the size of the microcarrier, then one would expect that the eddy would dissipate its energy against the surface of the microcarrier and thereby cause damage to the attached cells (Croughan et al., 1987, 1989; Cherry and Papoutsakis, 1986, 1988). The energy contained in the eddies is transferred from large to small eddies. The energy of the resulting small eddies is then dissipated as heat. The length scale of these smallest eddies, which are the most detrimental to cells on the basis of the above argument, is given by the Kolmogorov length scale (Tennekes and Lumley, 1985), which is a function of power dissipation per unit mass and kinematic viscosity. I t has been demonstrated that cell damage in microcarrier cultures begins to occur when the Kolmogorov length scale is comparable to the microcarrier diameter and that a decreasing Kolmogorov length scale leads to increasing damage to these cells (Croughan et al., 1987,1989; Cherry and Papoutsakis, 1986,1988). Croughan et al. (1989) also
0 1990 American Chemical Society and American Institute of Chemical Engineers
392
demonstrated that increased viscosity, which results in an increased Kolmogorov length scale, reduced cell damage. These results demonstrate that damage to cells on microcarriers correlateswell with the Kolmogorov length scale, in terms of both power dissipation per unit mass and kinematic viscosity. It is unclear whether this concept of cell damage caused by turbulent eddies can be extended to cells growing in suspension. In capillary tube experiments performed by McQueen et al. (1987),the specific cell lysis rate correlated well with the Kolmogorov length scale,becoming significant when the estimated Kolmogorov length scale was below 3.5 pm. The average cell size was approximately 10 pm, with some cells as small as 5 pm. This result is consistent with the hypothesis that cell damage occurs when the smallest eddy size is comparable to the size of the cells. Subsequent results by McQueen and Bailey (1989), however, did not show such a correlation between the eddy size and cell damage. In these experiments, a 50 % increase in the viscosity (which results in a 36% increase in the Kolmogorov eddy length) had no effect on the specific cell lysis rate. In this study, the effect of agitation on cell growth was investigated by using the attachment-independent Spodoptera frugiperda Sf9 insect cell line as a model system. The mechanism of “bubble damage” and Pluronic F-68 protection in sparged bioreactors was also investigated. Two airlift bioreactors, which differed principally in their sparger design, were considered. The severity of bubble damage in one of the airlift bioreactors permitted comparison of the protective effect of Pluronic F-68 with another Pluronic polyol, L-35.
Materials and Methods Cells and Medium. The cell line used in this study, S. frugiperda Sf9 insect cells, was generously provided by Max Summers, Department of Entomology, Texas A&M University. These cells are also available from the American Type Culture Collection (cat. no. CRL 1711). The cells were maintained in 25- and 75-cm2tissue culture flasks (Corning Glass Works, Corning, NY). Cells grown in 50-mL spinner flasks (Model 1967-00050, Bellco Biotechnology, Vineland, NJ) at 27 OC and 100 rpm were used to seed the spinner flasks and bioreactors. Cells were grown in TNM-FH medium (Summers and Smith, 1987) supplemented with 50 pg of gentamycin sulfate/mL (Sigma Chemical Co., St. Louis, MO), 2.5 pg of Fungizone/mL (Gibco Laboratories, Grand Island, NY), and 5% heatinactivated (45 min at 56 “C) fetal bovine serum (Hyclone, Logan, UT). The osmolality and pH of the medium were 315 f 10 mmol/kg and 6.2 f 0.1, respectively. Commercial-grade Pluronics F-68 and L-35 were generously provided by the BASF Corporation (Parsippany, NJ). Dow Corning medical-grade antifoam C was generously provided by Dow Corning Corporation (Midland, MI). Solutions of Pluronics F-68 and L-35 and antifoam C were sterilized by autoclaving (15 min at 121 OC). In the cases when medium was supplemented with a Pluronic poly01 or antifoam C, the concentrations (unless stated otherwise) were 0.241 (w/v) and 20 ppm, respectively. Cell Counts and Population Doubling Time. Cell counts were measured with a Coulter multisizer. All particles in the range of approximately 9-25 pm were considered as cells. Since there was some minor cell clumping, the numbers obtained with the Coulter multisizer are probably slightly low. The population doubling times were calculated from a least-squares fit of the growth curve data (counts from Coulter multisizer) in the exponential growth region.
Biotechnol. frog.., 1990, Vol. 6, No. 5
Agitated 3-LBioreactor and Airlift Bioreactors. A 3-L, water-jacketed, agitated bioreactor was used (Applikon, Foster City, CA) with a working volume of approximately 1.5 L. This bioreactor was agitated with either a flat six-blade impeller (diameter = 4.5 cm) or a scoping marine impeller (diameter = 4 cm). A scoping impeller results in fluid lift along the central axis in the bioreactor. The diameter of the bioreactor is 13 cm. At the usual operating volume, the fluid depth was also approximately 13 cm; therefore, the H I D (height/diameter) ratio was 1. The bioreactor contained four baffles, each 1.3 cm wide (i.e., 0.1D). The bioreactor was oxygenated by surface aeration with air/oxygen to maintain the dissolved oxygen concentration at a value between 50 and 100% of air saturation. It has previously been determined that cell growth is not affected by varying the dissolved oxygen setpoint from 50 to 100% of air saturation (Murhammer, 1989). Air was supplied to the bioreactor with an air pump. The oxygen was extra dry grade (Linde). The gases were sterilized by passage through a Millex-FGao 0.2-pm filter unit (Millipore Corp., Belford, MA) prior to introduction into the bioreactor. Due to the effectiveness of the bicarbonate buffer system in the medium, pH control was not necessary. Typically, the pH varied less than 0.1 pH unit over the course of a cell growth experiment. Disposable airlift bioreactors (Cellift model) were generously provided by Ventrex Laboratories (Portland, ME). The diameter and fluid depth in this bioreactor were approximately 6 and 27 cm, respectively. Thus, the H I D ratio was approximately 4.5. The gas was introduced into the bioreactor through a thin membrane gas distributor (pore size C0.2 pm), which also acts as a sterile filter. The air flow rate supplied to the airlift was 2.2 f 0.2 mL/ min., corresponding to a line pressure of 1.4 f 0.2 psi (as determined by the pressure gauge provided with the Cellift). This sparging rate corresponds to approximately 0.0029 VVM (volume of sparged gas per volume of reactor per minute). The air (Linde breathing grade) was provided from a gas cylinder. An autoclavable airlift bioreactor was purchased from Kontes (Vineland, NJ). The diameter and fluid depth in this bioreactor were approximately 5 and 35 cm, respectively. Thus, the H I D ratio was approximately 7. Air was passed through a Millex-FGao 0.2-pm filter unit (Millipore Corp.) prior to introduction into the bioreactor. Two different porous stainless steel gas distributors were used, with average pore sizes of 5 and 20 pm. The air flow rate was supplied to the airlift at 2.2 f 0.2 and 2.0 f 0.3 mL/ min (corresponding to 0.0032 and 0.0029 VVM, respectively) through the 5- and 20-pm distributors, respectively. These flow rates correspond to line pressures of 2.2 f 0.3 and 2.9 f 0.3 psi (as determined by the pressure gauge provided with the Cellift) through the 5- and 20-pm distributors, respectively. The air (Linde breathing grade) was provided from a gas cylinder. Bubble Diameter and Pressure Drop Across Gas Distributors. Bubble size was estimated by using a modification of the photographic method (Miyahara et al., 1983; Voyer and Miller, 1968). Briefly, a scale with l-mm graduations was placed within the respective bioreactor (containing only medium, i.e., without cells) and pictures were taken of the bubble flow with slide film. The resulting slides were then projected and the arithmetic average bubble sizes determined. The pressure drops across the gas distributors were calculated by subtracting the pressure drop due to the liquid in the respective bioreactor from the measured line pressure (as determined by the pressure gauge provided with the Ventrex airlift bioreactor).
393
Biotechnol. Prog., 1990, Vol. 6, No. 5 10’
-t-
0.2%Pluronic F-68
io4 0
25
50
75
100
125
150
175
200
Hours in culture
Figure 1. Typical results demonstratingthe effect of cavitation on the growth of S.frugiperda Sf9 insect cells. These experiments were performed in the Applikon 3-L bioreactor agitated at 850 rpm with a scoping marine impeller.
Calculation of Kolmogorov Eddy Length. The calculation methodology is given in detail elsewhere (Murhammer, 1989). Briefly, the Kolmogorov eddy length (7)was calculated as follows (Tennekes and Lumley, 1985): 7 = (V3/€)1’4 (1) where u is the kinematic viscosity and E is the power dissipation rate per unit mass. The kinematic viscosity was assumed to be the same as that of water. The power dissipation rate per unit mass was calculated from
= Nf13D: (2) where Npis the power number, N is the impeller rotational speed, and Di is the impeller diameter. Equation 2 is based on the assumption (Placek and Tavlarides, 1985) that energy dissipation is greater near the impeller; therefore, Di3 is assumed to be the volume into which the energy is dissipated. Np was obtained from the appropriate power number curve (function of Reynolds number and impeller type) given in Figure 15 of Rushton et al. (1950). t
Results Effect of Agitation on Insect Cell Growth. First, the effect of mild agitation on insect cell growth was investigated. These studies were performed in medium supplemented with 20 ppm antifoam C (except in the 25cm2 tissue culture flasks) in the absence of a protective agent (i.e., without Pluronic F-68). The cells grew equally well in 25-cm2tissue culture flasks, in 50-mL spinner flasks agitated at 100 rpm, and in the 3-L bioreactor (oxygenated by surface aeration) agitated at 200 rpm with a scoping marine impeller. The population doubling times in the tissue culture flasks, 50-mL spinner flasks, and 3-L bioreactor were 19.8 k 2.1 h (95?6 confidence level of the mean based on five growth curves), 18.4 f 0.7 h (95% confidence level of the mean based on 17 growth curves), and 21.6 f 3.3 h (95% confidence level of the mean based on six growth curves), respectively. These results demonstrate that mild agitation has no adverse effects on the growth of these insect cells. Next, the effect of higher agitation rates on cell growth was investigated. These studies were performed in the 3-L bioreactor with either a flat six-blade impeller or a scoping marine impeller. The bioreactor was oxygenated by surface aeration in order that the effect of agitation could
be uncoupled from the effect of sparging. An agitation rate of 300 rpm with the flat-blade impeller had no adverse effect on cell growth. The population doubling times were 18.8 and 20.0 h in the bioreactor and in the spinner flask control (100 rpm), respectively. These agitation rates correspond to Kolmogorov eddy lengths of approximately 25 and 85 pm in the bioreactor and spinner flask, respectively (Murhammer, 1989). Similarly, an agitation rate of 500 rpm with a scoping marine impeller had no adverse effect on cell growth. The population doubling times were 21.7 and 20.0 h in the bioreactor and in the spinner flask control (100 rpm), respectively. The corresponding Kolmogorov eddy length was approximately 38 bm in the bioreactor (Murhammer, 1989). The Kolmogorov eddy lengths of 25 and 38 pm for agitation rates of 300 rpm (flat-blade impeller) and 500 rpm (scoping marine impeller) in the bioreactor, respectively, are significantly larger than the cell diameter of 1015 pm (Murhammer, 1989). It was not possible to achieve smaller Kolmogorov eddy lengths with higher agitation rates since this resulted in the introduction of bubbles into the bioreactor, via either the cavitation or vortexing. Effect of Bubble Incorporation via Cavitation on Insect Cell Growth. The effect of cavitation on cell growth was investigated in the Applikon bioreactor agitated at 850 rpm with a scoping impeller (Figure 1). The bioreactor was oxygenated by surface aeration. No Pluronic F-68 was present in the medium. Cavitation was not apparent at the initiation of agitation but began to develop within 1 h into the run. By 20 h into the run, cavitation was fully developed with many small bubbles present (mostly smaller than 1mm in diameter). Rapid cell death occurred upon initiation of cavitation in the absence of Pluronic F-68. This experiment was repeated with medium supplemented with 0.2 % (w/v) Pluronic F-68. Cavitation did not begin to develop until 66 h into the run. By 91 h into the run, fully developed cavitation was present. By this time, cell growth was in transition between exponential growth and stationary phase (Figure 1). Medium supplemented with 0.2% (w/v) Pluronic F-68 provided essentially full protection from the detrimental effects of cavitation (population doubling time of 23.0 h vs 20.3 h in the spinner flask control). A foam layer 1-2 cm thick formed on the medium surface both with and without Pluronic F-68 added to the medium. Effect of Bubble Incorporation via Vortexing on Insect Cell Growth. The effect of an agitation rate of 800 rpm with the flat-blade impeller was investigated. This agitation rate resulted in severe turbulence within the culture and incorporation of bubbles via vortexing. The introduced bubbles were rapidly broken into very small bubbles (