Biotechnol. Prog. 1995, 1 I , 584-588
584
NOTES Partial and Total Cell Retention in a Filtration-BasedHomogeneous Perfusion Reactor Gautam G. Banik and Carole A. Heath*,+ Thayer School of Engineering, Dartmouth College, Hanover, New Hampshire 03755
Suspended mammalian cells can be cultivated in a variety of operational modes (pure chemostat, total cell retention, or partial cell retention) in a homogeneous perfusion bioreactor by varying the cell bleed rate. Hybridomas were grown in the reactor at a perfusion rate of 2.0 day-' for over 10 weeks at different specific growth rates and viable cell densities achieved by varying the extent of cell retention. Cell metabolism in the reactor was found to vary with the extent of cell retention, which determined both cell density and specific growth rate. With partial cell retention, the nutrient consumption and metabolite production rates decreased with both increasing growth rate and increasing cell density. The specific and volumetric antibody production rates, however, increased dramatically with cell density (and to a lesser extent with decreasing growth rate). The specific MAb production rate was lower with total cell retention than with partial retention at the same growth rate. Since the reactor can be operated over a range of perfusion rates and extents of cell retention, the system can be used to culture cell lines with widely different productivity patterns.
Introduction Continuous, homogeneous perfusion reactors are gaining favor as a means for cost effective production of proteins. Perfusion was first achieved by incorporating spin filters into a standard tank (Schmialeket al., 1976). The filter, which normally has a pore size larger than the cell diameter, retains suspended cells by means of tangential shear and the lift effect while removing spent medium. More recent modifications include changes in the method of cell retention; spin filters have been replaced by an external centrifuge (Tokashiki et al., 19901, an external tangential flow system (Jager et al., 1989; Seamans and Hu, 1990),or a cell settling tube (Batt et al., 1990). Another homogeneous perfusion system is the transtubular reactor, a modification of which is used in this study; porous tubes are used for transport of nutrients, wastes, and oxygen (Ozturk et al., 1989). Other filtration-based perfusion systems have also been developed (Graf and Schugerl, 1991; Buntemeyer et al., 1992; Hiller et al., 1993). In a previous study we demonstrated the effects of increasing cell density on hybridoma metabolism with total cell retention (Banik and Heath, 1994). To avoid the continuous accumulation of nonviable cells found with total cell retention, the reactor has been modified to allow partial cell retention by controlling the ratio of the rates of cell-free and cell-containingmedium removal. We report here the mass transfer characteristics of the modified system and demonstrate the reactor's ability to separately investigate the effects of cell density and specific growth rate on hybridoma metabolism and specific MAb productivity.
* To whom correspondence should be addressed.
Present address: Department of Chemical Engineering, Iowa State University, Sweeney Hall, Ames, IA 50011-2230. +
Materials and Methods Cell Culture. The hybridoma cell line H22 and the medium used in this study have been described previously (Banik and Heath, 1994). Perfusion Reactor. The system consists of a 1.2 L jacketed glass bioreactor with a working volume of 550 mL, maintained with a level controller (Cole Parmer). Four 1.5 m lengths of microporous poly(propy1ene) fiber (Akzo, Germany), uniformly coiled inside the reactor, were used for medium exchange. The fibers have an internal diameter of 1.8 mm, are 900 pm thick, have a pore size of 0.22 pm, and can withstand a transmembrane pressure up to 3 bars. Three fibers were used to remove cell-free medium from the reactor at any given time; the fourth fiber, alternated every 4-5 days, was back-flushed with fresh medium. The hydrophobic poly(propy1ene) membranes must be hydrophilized prior to use as described previously (Banik and Heath, 1994). The reactors were mixed with magnetic stirrers at 7075 rpm which also served to minimize the buildup of cells and cell fragments on the membrane surfaces. Three meters of hydrophobic poly(propy1ene)tubing was used to keep the dissolved oxygen concentration between 40 and 50% of air saturation by varying the 02,air, and CO, concentrations. The pH was controlled about a set point of 7.2 by supplying 5% COZ in the gas phase and by adding 0.5 M NaHC03 as needed. The reactors were inoculated from an exponentially growing batch culture and were run in batch mode until the cell density increased t o more than lo6 cells/mL, after which perfusion was initiated. A continuous flow of medium into the reactor determined the perfusion rate, D (feed rate divided by the reactor volume). The outlet stream consisted of two parts: cell-free broth pumped through the microporous fibers and cell-containingbroth removed directly from the
8756-7938/95/3011-0584$09.00/00 1995 American Chemical Society and American Institute of Chemical Engineers
Biofechnol. Prog., 1995, Vol. 11, No. 5
585
.
',ok
Q
. -
1
I
81
-E
Perfusare
Cell bleed
0.8
E 3
0
Surface aeranon Siliconc tubing Hollow fibn
90
IO0
6 0
f 0
1
2
3
4
e
80
110
Agitation Speed (rpm)
Fi ire 1. Residence time distributions as measured in th two streams exiting the reactor compared to a theoretical per ctly mixed reactor (solid line).
reactor, defining the cell bleed rate, B,. The ratio a = BJD indicates whether the reactor operates as a CSTR (a= 11, a total cell retention system (a= O), or a partial cell retention system (0 < a < 1). Analytical Procedures. Viable cell concentration and percent viability were determined by an average of six counts in a hemocytometer using erythrosin B. Glucose and lactate were measured using a glucose/ lactate analyzer CySI Model 2000, Yellow Springs, OH). ELISA was used to determine antibody concentrations. Glutamine concentration was measured using HPLC of the o-phthaldialdehyde derivatives using a (2-18column with gradient elution. An ion selective electrode (Model 95-12,Orion, Boston, MA) was used to measure ammonia. Determination of Kinetic Parameters. Viable (x,) and dead cell ( x d ) balances around the reactor yield the following equations:
where k d is the death rate. At steady state apparent specific growth rate, equals B,, and
70
60
50
40
5
papp,
the
(3) where p is the true specific growth rate and X is the total cell density. The metabolic quotients were determined as described previously (Banik and Heath, 1994). Reactor Characterization. Mixing in the reactor was characterized by residence time distribution (RTD) functions. An acid (0.1 N HC1) pulse input was applied at the inlet of the reactor, and the response was recorded using a flow-through pH probe in the cell bleed and perfusion streams and inside the reactor. The pH values were converted into H+ concentrations which were used to determine the RTD function E(@. The mass transfer coefficients for oxygen were measured by a standard dynamic technique modified for liquid flow. NZwas first passed through the fibers and the reactor head space. When the 02 concentration in the medium was constant, the gas was switched from NZ to air and the changing 02 concentration recorded.
Results Reactor Characterization. The perfusion reactor can be characterized as a perfectly mixed reactor (Figure
Figure 2. Oxygen mass transfer coefficient as a function of agitation speed for different types of aeration in the reactor. IW
90 80
70 60
E x
50
5
- I l l
'
1
Density Viabitiy
-C
' 0
~
40
~
,
.
.
,
.
,
.
,
.I
10
.
I
0
Time (hrs)
Figure 3. Viable cell density and percent viability in the reactor a t a perfusion rate of 2 day-' with partial cell retention.
1). Samples from the cell bleed and perfused streams were nearly identical to the perfectly mixed case except at very low residence times, suggesting that a small amount of fluid bypassed directly from the inlet to the cell bleed stream and/or the tubing. Experiments also demonstrated (not shown) that the E(@ curves are independent of a (BdD). Compared to surface aeration or silicone tubing, oxygenation by polytpropylene) fibers greatly improves mass transfer to the liquid (Figure 2). Air was passed into the reactor head space (650mL) at a flow rate of 1000 mL/ min for surface aeration alone; all air flow rates were equivalent. The increase in the mass transfer coefficient, with an increase in the agitation rate, indicates that, for the agitation speeds investigated, the liquid side mass tranefer resistance is limiting. Chemostat Operation. To determine whether the presence of fibers alters reactor performance compared to a CSTR, the perfusion reactor was operated in chemostat mode (a= 1)with and without fibers and the steady state cell densities as a function of dilution rate were compared. Except for washout, the two curves were the same (not shown). Washout occurred at dilution rates of approximately 1.25 and 1.35 day-l for the reactor without and with fibers, respectively, probably because of a small amount of cell settling on the fibers. Partial Cell Retention. With partial cell retention, the highest X, resulted at the lowest B, (Figure 3). A decrease in papp(equal to B,), however, had the effect of decreasing cell viability. Although the viability dropped to 70% at B, = 0.1 day-', X, remained constant at 1.20 x lo7 cells/mL. The lactate concentrations in the reactor remained within the range 28-36 mM for all bleed rates (Table
Biotechnol. Prog,, 1995, Vol. 11, No. 5
586
Table 1. Steady State Metabolic Parameters as a Function of Cell Bleed Rate at a Perfusion Rate of 2.0 day-' a parameter B, = 0.0 (day-') B, = 0.1 (day-') B, = 0.5 (day-') B, = 0.75 (day-') B, = 1.0 (day-') 0.14 0.89 1.13 0.63 specific growth rate (day-l) 0.12 0.83 1.42 2.78 5.56 glucose concn (mM) 0.64 34.67 34.4 32.78 30.22 36.4 lactate concn (mM) 4.03 4.14 5.70 9.72 4.10 qglu (mmol of gld(109cellsday)) 5.78 6.04 8.40 15.11 6.21 qlac (mmol of lac/(109cellsday)) 1.43 1.46 1.48 1.55 1.49 Y'Iadglu (mmol of gldmmol of lac) 0.06 0.18 0.16 1.32 0.03 glutamine concn (mM) 3.41 3.62 4.25 3.60 ammonia concn (mM) 3.81 1.00 1.5 2.33 1.01 1.02 qgin (mmol of gld(lO9 cellsday)) 0.59 0.63 1.09 1.80 q N H 3 (mmol of NH3/(109cellsday)) 0.66 0.59 0.62 0.73 0.77 0.64 Y ' N H ~(mmol ~ I ~ of gldmmol of NH3) 23.5 22.6 11.9 4.8 20.7 volumetric antibody prodn rate @g/(mL.h)) a The reported values for total retention (B, = 0.0 day-I) are at a quasi-steady state since, although viable cell density stabilized, the cell viability continued to drop.
1). Although lactate has been shown to inhibit cell growth at concentrations above 33 mM (Hassell et al., 19911, the cell viability remained at 80% or greater in the reactor for B, above 0.1 day-l. At B, = 0.1 day-', glucose was completely exhausted and the viability dropped to about 70%. Glucose consumption (Figure 4a) can be described by a typical maintenance energy model. Both qglu(consumption) and qlac (production) decrease with increasing X , (Figure 4b), as has been observed previously (Shirai et al., 1988, 1992; Ban& and Heath, 1994). Although X , was typically 1order of magnitude greater than that in a CSTR, the NH3 concentration remained between 3.4 and 4.2 mM, which is within the normal range (Hassell et al., 19911, for all bleed rates investigated (Table 1). Glutamine concentrations were low (0.05-0.15 mM) in the cell bleed range of 0.1-0.75 day-I. The specific rates of glutamine consumption and NH3 production increased with increasing p and decreased with increasing X , (Figure 4). With increasing X , and decreasing p, both of the apparent yield coefficients Y'ladglu and Y"Hdgln decreased steadily (Table 1).A decrease in Y'ladglu with decreasing ,D has also been observed in low-density chemostat cultures (Hiller et al., 1991). Without the drop in these apparent yield coefficients under conditions of low p and high X,, perfusion culture would be less attractive as a culture method because of the potentially toxic lactate and ammonia concentrations. Total Cell Retention (B, = 0). The reactor was run with partial cell retention at D = 2.0 day-l and B, = 1.0 day-l (-100 h); when X , stabilized, the reactor was changed to total retention (0h). Although X, remained high with total retention, the viability dropped considerably (Figure 5). Although glucose in the reactor was not completely exhausted, it reached a very low level (0.65 mhQ, and the glutamine level was essentially zero at the end of the run (Table 1). The concentrations of glucose, glutamine, and viable cells were steady for the final 100 h of culture, but the viability continuously decreased as the lactate and NH3 levels slowly increased. Thus, although papp during this period was almost zero (-0.02 day-l), p was 0.121 day-' and cell metabolism was not steady, as evidenced by slowly increasing levels of lactate and NH3 (not shown). The final concentrations of lactate and NH3 were slightly higher than those at the same p (0.141 day-') in partial cell retention culture at D = 2.0 day-' and Bc = 0.1 day-l. During the initial 200 h, the specific rates of glucose and glutamine consumption and the specific rates of lactate and ammonia production all decreased with time, but after X , peaked and the viability started to decrease, the metabolic quotients increased slightly. The increase
-
.&
. -I t
161 '
12-
a
Glucose consumption Lactateprcducdon 0 Glutamine conrumpnon m Ammoniapmduction 0
