Oxygen transfer rates in a mammalian cell culture bioreactor equipped with a cell-lift impeller

Oxygen Transfer Rates in a Mammalian Cell Culture Bioreactor Equipped with a Cell-Lift Impeller

M. Johnson,* G. Andre, C. Chavarie,* and J. Archambaultt National Research Council Canada, Biotechnolog y Research Institute, Montreal, Canada H4P 2R2

Accepted for publication April 3, 1989

Measurements of k,a were carried out in 1.5- and 5-L New Brunswick Scientific CelliGen@ bioreactors. The measured kLa in water were identical for both vessel sizes operated in similar condition. The mass transfer rate increased with temperature, mixing speed , and aeration rate, with this last parameter being the most significant. Surface aeration alone gave kLa values of 0.4 to 1.6 h-‘. A 25% decrease in k,a was observed above an aeration rate of 1.6 vvrn. This was caused by the par- ticular foam breaker of t he CelliGen bioreactor. Mea- s u r e m e n t s of kLa using a mammal i an cell cu l tu re medium supplemented with 5% fetal calf serum (FCS) have confirmed the negative effect of the foam breaker on kLa. The measured value in this medium was 1.2 h-’ for all aeration rates, more than 60% of which was attributed to surface aeration.

INTRODUCTION

The design criteria of a mammalian cell bioreactor include supplying efficiently sufficient oxygen to the culture for survival and growth of the cells and product biosynthesis. This function must be performed under minimal stressing conditions to the fragile cells while meeting their significant demand for this essential nutrient. A typical range of oxygen uptake rates (OUR) for mammalian cells is of the order of 0.05 to 1.0 mmol O2 L-’ h-’ at lo6 cells/mL, which requires oxygen transfer capacities (kLa coefficient) of 0.25 to 5 h-’ for the culture system.’

The gas exchange impeller of the CelliGen bioreactor marketed by New Brunswick Scientific (NBS) was de- signed to meet these requirements. It features a multipur- pose impeller (Fig. 1) combining aeration of the culture medium within a rotating cage, collapsing of the foam into the upper portion of the cage, and macrocirculation mixing patterns of‘ the liquid phase generated by the hollow central shaft fitted with three rotating horizontal jet tubes. Accord- ing to the manufacturer, this device should allow adequate culture aeration while preventing direct contact between

* Present address: !&ole Polytechnique de MontrCal, Department of Chemical Engineering, Montreal, Canada H3C 3A7.

+ To whom correspondence should be addressed.

Biotechnology and Bioengineering, Vol. 35, Pp. 43-49 (1990) 0 1990 John Wiley & Sons, Inc.

cells and air bubbles and minimize foam formation without antifoam addition. Furthermore, adequate low shear mixing is ensured by the hydrodynamic pattern generated by the rotation of the impeller resulting in bulk movement of cells and media (referred to as the “cell-lift effect”).

The kLa of this bioreactor (5-L vessel, 50 rpm) was reported to vary from 1.5 to 5.5 h-’ when sparging air (10.625 vvm) into the cage and from 0.4 to 1.5 h-’ with surface aeration alone.* These values contrast markedly with those expected for conventional microbial fermentors (50-1000 h-1),3 and could still be limiting in regards to the appreciable oxygen requirements of actively growing and/or high density (> lo7 cells/mL) mammalian cell cultures.

This study was carried out to evaluate the oxygen transfer capacity of the CelliGen bioreactor for aeration and agitation rates ranging to 2 vvm and 100 rpm, respectively, in tap water and in a typical culture medium supplemented with 5% fetal calf serum (FCS). The influence of the temperature, agitation, and aeration rates on kLu was determined and the effect of the foam breaker on the operation and oxygen transfer capabilities of the bioreactor was assessed.

MATERIALS AND METHODS

Bioreactor

Measurements of kLa were performed in the NBS CelliGen bioreactor shown in Figure 1. Glass vessels of 1.5 and 5.0 L were used with respective working volumes of 1.2 and 3.5 L, internal diameters of 11.9 and 15.3 cm, and liquid heights of 14.4 and 21.0 cm. Mixing and aeration were achieved using the magnetically entrained gas exchange impeller positioned at 5 mm from the bottom of the vessel. The pH and dissolved oxygen probes and the medium sam- pling system were installed to generate the hydrodynamics representative of the normal bioreactor configuration. The inlet gas flow rate was measured by an external rotameter (Linde model No. UCCP/N FM 4344) and verified for leaks at the outlet with a wet test gas meter (Precision Scientific, model 6315).

CCC 0006-3592/90/01043-07$04.00

GAS IN

ELIMJNATOR

I 1 '1 DR;VE MAG!:€ 1

Figure 1. Geometrical configuration of the CelliGen bioreactor.

The CelliGen bioreactor is equipped with an interactive system for gas addition (air, 02, N,, CO,) to control the dissolved oxygen concentration (DO) and pH. This system operates in a cyclic mode, the gas mixture being sparged into the bioreactor for 3 min and turned off for 0.5 min, to dislodge cells attaching on the impeller external surface as a result of liquid flow through its ~ c r e e n . ~ This control system was bypassed during k,a measurements to avoid unde- sirable fluctuations of the incoming gas flow rate. The only gases used were air for absorption experiments and nitrogen for stripping oxygen from the liquid phase.

Measurement of Dissolved Oxygen Concentration

The concentration of oxygen dissolved in the liquid phase was measured with a galvanometric oxygen electrode (NBS model 800) and recorded on a LKB Bromma model 2210 chart recorder. The calibration of this probe was affected by the liquid phase temperature, pressure, and hydrodynamic regime in the vicinity of the probe. Calibration of the probe was done at 175 rpm, at atmospheric pressure, and at the operating temperature. Nitrogen or air was sparged directly

into water at 12 L/min (STP) until a constant DO was observed (0% or 100% of saturation). The meter and recorder were adjusted accordingly.

K,a Measurements

A number of methods are available' to measure k,a, based on the sulfite oxidation technique, on global oxygen balance, on the steady-state continuous flow technique, or on the dynamic absorptionidesorption technique. The latter was used in this study because it is convenient, accurate, and easy to implement at the bench scale. It consists of monitoring the response of the dissolved oxygen concentration following a step change in the inlet gas composition, i.e., from nitrogen to air.

The dynamic response of the DO probe can affect the calculated values of k,a if the system characteristic time for mass transfer (l/k,a) is of the same order as the response time of the probe. Assuming that the probe dynamics fol- lows a first-order model characterized by a time constant T,, , a simple criterion for neglecting this effect is T~ l/k,a.' This T, was measured by transferring the probe from a

44 BIOTECHNOLOGY AND BIOENGINEERING, VOL. 35, JANUARY 1990

nitrogen-sparged solution to an air-saturated solution. This time constant was found to be 12 s, which is significantly smaller than the lowest mass transfer characteristic time observed (>330 s). Consequently, the probe dynamics could be neglected when calculating the kLa values.

The hydrodynamic regime in the vicinity of the probe can also affect the measured DO. Depending on the thickness of the boundary layer next to the probe membrane, a concentration gradient will be established outside of the probe even when steady state is reached. For instance, during these experiments, the probe signal at saturation decreased with the impeller speed, ranging from 100 recorder chart units at high speed (175 rpm) to approximately 85 chart units at 50 rpm. The additional static and dynamic resistance offered by the boundary layer would slow down the dynamic response of the probe leading to smaller kLa values. However, it is estimated that the boundary layer may contribute an additional resistance equivalent to about one-third to one-half of the resistance offered by the probe membrane itself, which was shown previously to be negligible compared to the mass transfer dynamics of the system. Finally, the effect of the gas phase dynamics was neglected owing to the small values of kLu measured (0.5-11 h-I).

The k,a value was calculated using the following unsteady- state mass balance equation performed on the liquid phase:

which, upon integration of (C* = C,), yields

Nitrogen was sparged into the reactor until complete oxygen desorption was achieved [CL(0) = 01. The value of C,(t) was continuously recorded following the step change in the inlet gas composition (switching from nitrogen to air). Since CLr is a function of the mixing speed, it was determined for each experience as the stable recorder read- ing when the absorption was completed. The kLa value was calculated by a linear regression of the recorded data transformed as per eq. (2).

Operating Conditions and Culture Medium

Most kLa measurements were performed in tap water at 27 and 37"C, at agitation speeds of 25 to 100 rpm, and at aeration rates of 0.14 to 2 vvm. Some measurements were also performed using a typical mammalian cell culture medium made of a 1 : 1 mixture of Fisher's medium and Dulbecco's Modified Eagle Medium (DMEM) which was filter steril- ized (ACROSO No. 4401 0.2-pm hydrophobic filter). This basal medium was supplemented with 5% (v/v) fetal calf serum and introduced into the sterile bioreactor for experi- mentation. All medium components were purchased from Gibco Laboratories.

RESULTS

Effect of the Mixing Rate on k,a

The oxygen transfer coefficient of both bioreactors was measured using tap water at 27 and 37°C at mixing speeds ranging from 25 to 100 rpm and at aeration rates as indi- cated in Table I . The results are presented in Figure 2 and correlated in Table I . The kLu values showed a weak linear dependency on the agitation speed.

Effect of the Aeration Rate

Results of the effect of the aeration rate on kLa in both bioreactors operated at 50 rpm and at 27 and 37°C are illus- trated in Figure 3 and correlated in Table I . Surface aeration (corresponding to a sparged aeration rate of 0 slpm) was measured at a headspace gas renewal rate of 1.5 slpm. The kLa values for surface aeration ( 1 1 . 6 h-') which are presented in Table 11, were not included in the correlations of Table I .

The oxygen transfer coefficient of the 1.5-L bioreactor increased linearly with the aeration rate to 1.6 vvm. There- after it dropped by 25%. Simultaneously, the liquid level in the aeration cage of the gas exchange impeller was observed to decrease significantly (to 50% of the maximum value) as the air flow rate was raised (Table 111). This emp- tying of the aeration cage resulted in the negative effect observed on kLu.

Table I.

VL T Q N Correlation for Line (L) ("c) (L/L/min) (min-') kLa (h-') r Figure No.

Summary of kLa correlations in tap water.

1.2 37 1.2 27 3.5 37 1.2 3 1 1.2 21 3.5 37 1.2 37

1.2 27 3.5 37

3.5 37

1.25 1.25 0.43 0.42 0.42 0.14

0.15-1.6

0.15-1.6 0.15-1.6

0.15- 1.6

25- 100 25-100 25-100 25-100 25-100 25-100

50

50 50

50

kLa = 0.023N + 8.01 0.97 2 kLa = 0.014N + 7.51 0.97 2 kLa = O.007N + 5.18 0.63 2 kLa = 0.009N + 4.03 0.79 2 kLa = 0.02N + 2.90 0.97 2 kLu = 0.012N + 2.25 0.99 2 kLa = 5.00Q + 2.77 0.99 3

kLa = 4.35Q + 2.31 0.99 3 k,a = 5.00Q + 2.62 0.93 3

4 k,a = 7.010 + 2.26 0.99 4

5

1 2 3 4 5 6 1 1 3 2 2 1 -

JOHNSON ET AL.: OXYGEN TRANSFER RATES 45

11.0 r

r v

o 5.0 _1

Y

3.0

3'

4 0 /.---+---- /--

- _/---

5 t- n 9 P v

6 C -

v

10 30 50 70 90 110

Agitation speed (RPM)

Effects of agitation, aeration, and temperature on k,a in tap wafer: Figure 2. (A) 1.25 L/L/min, 37"C, VL = 1.2 L; (A) 1.25 L/L/min, 27°C. VL = 1.2 L; (0) 0.43 L/L/min, 3 7 T , V, = 3.5 L; (0) 0.42 L/L/min. 37°C. VL = 1.2 L; (*) 0.42 L/L/min, 27°C. V, = 1.2 L; (0) 0.14 L/L/min, 37"C, V, = 3.5 L.

12.0 1

10.0 -

8.0 - /-. 7

I 5 6.0 -

0.0 0 . 3 0 .6 0.9 1.2 1.5 1.8 2.1

Aeration rate (L/L/rnin)

Figure 3. Effects of aeration and temperature on kLa in tap water at 50 rpm. Values at Q = 0 correspond to surface aeration:

(0) VL = 3.5 L, 37°C; and (*) V, = 1.2 L, 27°C.

(0) v, = 1.2 L, 37°C;

Table 11. 1.5 L/min).

Values of kLa for surface aeration alone (aeration rate of

Tap water 3.5 27 27 37 37

Tap water 1.5 27 27 37 37

Medium 1.5 37 37

50 0.43 100 1.44 50 0.54

100 1.50 50 0.69

100 1.58 50 0.95

100 1.51 50 0.81

100 1.44

Effect of the Foam Breaker

In order to investigate the cause of this behavior which lim- ited the oxygen transfer capacity of the system, the foam breaker was removed from the impeller of the 5-L bioreactor. The liquid level in the aeration cage remained nearly constant (Table 111: 11.5-12.5 cm) for all aeration rates. As shown in Figure 4, k,a values were correspondingly increased by 25% and more above 0.6 vvm.

K,a Measurement in a Mammalian Cell Culture Medium

All previous experiments were performed using tap water as the liquid phase. In order to extrapolate these results


Table 111. Water level in the aeration cage (50 rpm, 37°C)

Aeration Water rate level

VL Foam breaker (L L-' min-') (cm)

1.5 Yes

3.5 Yes

3.5 No

0.42 0.83 1.25 1.67 2.0 0.14 0.43 I .oo 1.29 I .57 0.14 0.40 0.77 1.11 1.26 1.37

6.0 5.2 4.8 4.5 2.7

12.0 11.0 8.0 7.5 7.0

12.5 12.4 12.0 11.7 11.5 12.0

to a more realistic situation, the k,a of the 1.5-L bioreactor equipped with the cell-lift impeller (37"C, 50 rpm, 1-2 vvm) was measured in a culture medium supplemented with 5% FCS with and without the foam breaker. Results of this experiment are summarized in Figure 5 . The level of liquid in the aeration cage could not be measured because of the opacity of the medium.

The k,a of the bioreactor equipped with the foam breaker remained low (1.2 h-') for all aeration rates. Partial removal of the foam breaker improved significantly k,a (300%), which was found lower than measurements in tap water upon sparging air.

DISCUSSION

As shown in Figure 2 , the effect of the mixing rate on k,a is minimal. A marginal increase of less than 25% was

observed as the impeller speed was increased from 25 to 100 rpm regardless of the reactor volume, temperature, or aeration rate. This contrasts with conventional (Rushton) impellers which contribute to break up gas bubbles and increase significantly oxygen transfer to the liquid phase. In the case of the CelliGen bioreactor equipped with the cell- lift impeller, increasing the level of agitation results in better mixing, in better medium exchange between the aeration cage and the bulk of the culture volume and in improved surface aeration (Table 11: 0.5-1.5 h-I) , the contribution of which is not negligible under real culture conditions.

The results presented in Figures 2 and 3 show that kLa increases more rapidly with the aeration rate, from 3 to 11 h-' in the range of 0.14 to 1.6 vvm. These results are in good agreement with those of Reuveny et al.' who reported similar trends but somewhat smaller k,a values since they used a phosphate buffered solution (pH 7) as the liquid phase instead of tap water. The mass transfer rates were also found to be independent of the reactor volume as the results for both the 1.5- and 5.0-vessels were correlated with similar curves (Table I).

A reduction of temperature from 37 to 27°C decreased consistently the k,u coefficient by about 1 to 1.5 h-I. NO explanation was found for this phenomena. However, this has been reported elsewhere.'

The true mass transfer performance of the gas exchange impeller (k,a), can be evaluated by subtracting the surface aeration contribution (kLa), from the kLa values measured for the whole reactor and by multiplying this difference by a correction factor (V,/V,-) according to the following equation

(3)

This volume correction factor takes into account the fact that the measured k,a value was based on the total volume of liquid in the reactor (V,) whereas only the volume of

(k,a), = [k,a - (k,a),lvL/vc

0.0 0.3 0.6 0.9 1.2 1.5 1 .a Aeration rate (L/L/min)

Modification of the CelliGen configuration (3.5 L, 50 rpm, 37°C): Figure 4. (0) basic configuration with the foam breaker in place and (m) operation without the foam breaker.


12.0 1

10.0

8.0

6.0

n v--

1 I v

0

Y -I

-

-

-

4.0 i ,,i

\ \ \ \ \

Q l

A

0.0 0.3 0.6 0.9 1.2 1.5 1.8 2.1

Aeration rate (L/L/rnin)

Figure 5 . 50 rpm, 37°C). Values of k,a at Q = 0 correspond to surface aeration alone: (0) tap water, (0) culture medium with foam breaker, and (A)culture medium with partial removal of the foam breaker.

Values of k,a for the culture medium supplemented with 5% FCS (V, = 1.2 L,

liquid in the cage (V,) was subjected to gas-liquid transfer. This (k,a), coefficient is more representative of the effective mass transfer rate occuring within the aeration cage, which can be considered as a bubble column of cross- sectional areas of 11.66 and 12.80 cm*, respectively, for the 1.5- and 5.0-L vessels.

The (,$a), of both bioreactors are plotted against the gas superficial velocity based on the cage cross-sectional area in Figure 6. They are compared to typical bubble column correlations.8 The experimental data can be correlated by the usual equation

&,a), = aug (4)

As seen in Figure 6, the experimental results approaches the values obtained by Deckwer for a bubble column equipped with a sintered metal porous sparger (line 2). The calculated values of the a and /3 parameters of eq. (4) were 0.0242 s - ' and 0.92 for the 1.5-L reactor and 0.0217 s-I and 0.75 for the 5.0-L reactor. These are in good agreement with the values of a and /3 published by Deckwer for a porous sparger (0.0274 s-' and 0.80). Con- sequently, the aeration cage performs as an effective

I ' ' - 7

Figure 6. locity in tap water at 50 rpm and 37°C: (Line 1) Bubble column with multi-orifice sparger (ref. 8); (Line 2) Bubble column with sintered metal porous sparger (ref. 8); (0) V, = 3.5 ; and

Values of experimental (k,a), in the aeration cage as a function of air superficial ve-

(0) v, = 1.2 L.


bubble column. However, the experimental values of (k,a), and fitted coefficients seem to indicate some deterioration of the system performance upon scaleup from the 1.5-L to the 5.0-L bioreactor. This would likely be related to the fact that the two vessels are not geometrically similar. The working volume, external cage area and height ratios are 2.9, 2.1, and 2.0, respectively, while the bubbling cross sectional areas differ only by 10%. Thus, it is expected that the transport rate of oxygen by convection from the aeration cage to the bulk of the liquid phase will be rela- tively smaller in the larger bioreactor.

As shown in Figure 3 , the k,a values were found to drop by 25% above an aeration rate of 1.6 vvm. This decrease was attributed to a substantial reduction in the effective volume of liquid sparged in the aeration cage (Vc). In fact, increasing the aeration rate resulted in progressively higher pressure drops across the wire mesh screen used to break the foam (Fig. 1). This, in turn, led to the observed reduction in the level of liquid inside the aeration cage. This inefficiency of the aeration system was shown to be more pronounced when using a typical cell culture medium (Fig. 5 ) as the proteins contained in the liquid phase fouled rapidly and permanently the foam breaker. Consequently, oxygen transfer to the liquid phase under real culture conditions occurs mainly (60%) by surface aeration.

The cyclic mode of operation of the CelliGen gassing system described previously ( 3 min on and 0.5 min off) is not expected to improve the reactor's mass transfer rate significantly (c0.5 h-') because of the dynamics of this cycle and the extent of the foam breaker blockage. Higher gas pressure in the cage resulted in bubbling in the bulk liquid phase. The removal of the foam breaker resulted in constant level and V, in the aeration cage for all air flow rates. Modification of this foam breaker using, for ex- ample, a larger wire mesh screen and/or using a low serum or protein free culture medium may not improve the k,a of the system as either screen fouling by macromolecular compounds (cell debris, products, etc.) and/or foaming will occur, especially upon scaleup.

M. Johnson was a recipient of a Natural Science and Engineering Research Council of Canada Scholarship.

NOMENCLATURE

dissolved oxygen concentration at equilibrium and at saturation with air (mg L-I) dissolved oxygen concentration at time t (mg L-') diffusivity of oxygen in the membrane of the DO probe (cm2 s) overall volumetric oxygen transfer coefficient (h-l) volumetric transfer coefficient by surface aeration (h-l) volumetric transfer coefficient in the aeration cage (s-') DO probe membrane thickness (cm) agitation speed (min-I) aeration rate (L L-' min-I) temperature ("C) superficial air velocity in the aeration cage (cm sC1) total liquid volume (L) volume of liquid in the aeration cage (L) parameter of eq. (4) (s) parameter of eq. (4) DO probe time constant (s).

References

1 . R. Spier and B. Griffiths, Develop. Biol. Standard., 55, 81 (1984). 2. S. Reuveny, Z. B. Zhing, and L. Eppstein, Am. Biotechnol. Lab., 28

3 . M. Lavery and A. W. Nienow, Biotechnol. Bioeng., 30, 368 (1987). 4. New Brunswick Scientific, Instruction Manual for the CelliGen Biore-

actor, March 5, 1987. 5. M. Sobotka, A. Prokop., I. J. Dunn, and A. Einsele, Annual Reports

on Fermentation Processes Vol 5 , G . T. Tsao, Ed. (Academic, New York, 1982), p. 127.

6. C. W. Robinson, Ph.D. thesis, University of California at Berkeley, Berkeley, CA, 1971.

7. L. D. Benefield and C. W. Randall, Biological Process Design for Wastewater Treatment, &entice-Hall Series in Environmental Sciences (&entice-Hall, Englewood Cliffs, NJ, 1980), p. 281.

8. W.D. Deckwer, R. Burckhart, and G. Zoll, Chem. Eng. Sci . , 29, 2177 (1974).

(JadFeb 1986).


Documents

Oxygen transfer rates in a mammalian cell culture bioreactor equipped with a cell-lift impeller