Applied Microbiology and Biotechnology

, Volume 74, Issue 2, pp 324–330

Study of the oxygen transfer in a disposable flexible bioreactor with surface aeration in vibrated medium

Authors

    • Laboratoire Génie des Procédés Industriels UMR CNRS 6067, Département Génie ChimiqueUniversité de Technologie de Compiègne, Centre de Recherche de Royallieu
  • Jean-Michel Lebeault
    • Laboratoire Génie des Procédés Industriels UMR CNRS 6067, Département Génie ChimiqueUniversité de Technologie de Compiègne, Centre de Recherche de Royallieu
Biotechnological Products and Process Engineering

DOI: 10.1007/s00253-006-0682-1

Cite this article as:
Kilani, J. & Lebeault, J. Appl Microbiol Biotechnol (2007) 74: 324. doi:10.1007/s00253-006-0682-1

Abstract

The oxygen mass transfer is a critical design parameter for most bioreactors. It can be described and analyzed by means of the volumetric mass transfer coefficient KLa. This coefficient is affected by many factors such as geometrical and operational characteristics of the vessels, type, media composition, rheology and microorganism’s morphology and concentration. In this study, we aim to develop and characterize a new culture system based on the surface aeration of a flexible, single-used bioreactor fixed on a vibrating table. In this context, the KLa was evaluated using a large domain of operating variables such as vibration frequency of the table, overpressure inside the pouch and viscosity of the liquid. A novel method for KLa determination based on the equilibrium state between oxygen uptake rate and oxygen transfer rate of the system at given conditions was also developed using resting cells of baker’s fresh yeast with a measured oxygen uptake rate of 21 mg g−1 h−1 (at 30°C). The effect of the vibration frequency on the oxygen transfer performance was studied for frequencies ranging from 15 to 30 Hz, and a maximal KLa of 80 h−1 was recorded at 30 Hz. A rheological study of the medium added with carboxymethylcellulose at different concentrations and the effect of the liquid viscosity on KLa were determined. Finally, the mixing time of the system was also measured using the pH method.

Keywords

Surface aerationKLaGas–liquid mass transferPouchesBaker’s yeastVibration

Introduction

The development of aerobic cell culture in flexible, single-used pouches tends to spread in many applications using animal, vegetable cells and transgenic roots. These pouches have many advantages; they are disposable and pre-sterilized and therefore require no cleaning or in situ sterilization. They are equipped with ports for gas/liquid inlet and outlet. Measurement, additions and sampling are possible without the need for a laminar flow cabinet; these features dramatically lower the purchase cost and operating expenses of the system. The main problems encountered while using these pouches are essentially mixing and mass transfer; in fact, deprived of an agitation mobile, these systems must be fixed on an agitation platform animated with a swing or an orbital movement. Oxygenation is made by surface aeration, often by injecting air or oxygen through the headspace of the pouch (Singh 1999).

Sufficient oxygen supply for biological cultures is essential for efficient aerobic bioprocess development. According to Hilton (1999), often, in liquid batch processes, the critical parameter is gas exchange or balance between the oxygen demand of the culture (OUR, oxygen uptake rate) and oxygen transfer to the culture (OTR, oxygen transfer rate). The aeration efficiency of a given system is normally expressed by the volumetric mass transfer coefficient KLa. There are many methods for KLa measurement and determination, but not all of them are suitable when dealing with respiring and rapidly growing microorganisms or even in viscous fluids; the most used methods have been recently discussed and characterized (Gogate and Pandit 1999).

In the laboratory, devices such as spinner flasks, roller bottle, T flasks and similar systems are widely used for sensible culture such as animals and plant cells where bubbling and agitation can be critical parameters. However, these systems can only produce 100 to 1,000 ml of culture per batch due to their limited capacities of oxygen transfer (Hu et al. 1986).

Stirred tank bioreactor that are modified to reduce shear forces are still used for larger volumes, but they are still expensive, complex and unsuitable because of the use of bubble aeration and local fluid shear. Another culture system for the cultivation of animal, insect and plant cells has been developed by Wave Biotech. This system uses disposable pouches bioreactors fixed on a platform animated with a rocking motion. The agitation and oxygenation are induced by the wave propagation through the gas–liquid interface; the use of a similar idea for small-scale cell culture has been reported (Kybal and Sikyta 1985). In the wave bioreactor, the best KLa recorded was around 4 h−1, and the largest pouches tested have a capacity of 500 l (Singh 1999).

Our objective is to evaluate the OTR of a surface aerated bioreactor system using plastic pouches fixed on a vibrating table at high frequencies up to 30 Hz. Discussions taking into account OTR and OUR could give information on the types of cell cultures which can be performed with such a system.

Materials and methods

Microorganisms

Resting cells were prepared from commercial fresh yeast (Cappa, DSM Bakery Ingredients France S.A.S.). Fresh yeasts (4.5 g) were weighed, suspended in 50 ml of physiological serum (distilled water containing 0.9% NaCl) and then incubated at 30°C and 120 rpm for 15 min until the cells exhausted all of their nutritional reserves. Then, the solution is transvased in the pouch bioreactor containing 1.45 l of physiological serum added with 3 g l−1 of sucrose (the reference solution).

Fermentation unit

The fermentation unit used in this work consists of three main parts: a bioreactor, a vibrating table and a computer for on-line acquisition unit (Prelude, Pierre Guerin, France). The bioreactor was a heat-welded polyethylene bag performed at the laboratory starting from a plastic sheath of 150-μ thick. It had the dimensions of 400 × 420 mm and a working volume of 5 l. The bioreactor was equipped with different ports for gas inlet and outlet, sampling valve and housing for temperature probe and dissolved oxygen probe (DOP). The gas outlet tube is connected to a pressure indicator (PI, electronic presure cell) for in-pressure measurement as shown in Fig. 1.
https://static-content.springer.com/image/art%3A10.1007%2Fs00253-006-0682-1/MediaObjects/253_2006_682_Fig1_HTML.gif
Fig. 1

Representative scheme of the system

The air feeding flow was controlled by a precision gas mass flow controller (Brooks Instrument B.V.). The surface aeration is obtained by passing a continuous airflow of 0.15 l min−1 (0.1 vvm) through the headspace of the bag to provide oxygenation and gas exchange. Exhaust air passes through a closed glass bottle connected to the pressure indicator and a classic flow meter before being released. This system allows control of the in-pressure of the bag and prevention of overinflation and potential bursting.

The vibrating table model CTV 38/31 was purchased from Cassadio, Italy, and it has a variable frequency from 1 to 100 Hz.

Mixing and mixing time

Mixing time is defined as the time for the bioreactor to reach a new steady state after a step change in feed. There are many methods to determine the mixing time, among them are heat tracer, pH and conductance.

In this study, the pH method was used to determine the mixing time of the vibrating bioreactor. This method is sensitive and easy to carry out. The values of mixing time were determined by measuring the time required to attain a constant pH value after adding a pH effector (acid or base) to the bioreactor. The mixing time measurements were carried out by injecting 10 ml of a 3 N NaOH solution into a fermentor containing 1.5 l of distilled water (at an initial pH value of 7), and the resulting changes in pH were recorded and compiled into a curve by the computerized acquisition unit (Prelude). Ten milligrams of 3 N HCl was then added to the fermentor, and another pH curve was traced. The mixing time was estimated as the time needed to reach a constant pH value; an average mixing time was obtained from the two experiments. The experiments were carried out twice for each tested vibration frequency.

Oxygen measuring system

The system used for dissolved oxygen measurement and data acquisition consists of two parts: a steam sterilizable polarographic probe (Mettler Toledo) and a computer control and acquisition unit. The probe was maintained in a vertical position in the centerline of the bioreactor and immerged at half-depth of the liquid phase. The value of dissolved oxygen concentration (DOC) was recorded every 15 s and then plotted on-line against time by the computer unit. The DOP was calibrated to 100% DOC by flushing air through the liquid surface and maintaining the bioreactor at a fixed vibration frequency until reaching a steady state, then calibrated to 0% while disconnected. The DOP response time is less than 10 s.

KLa calculation

The volumetric oxygen coefficient KLa consists of KL, the mass transfer coefficient and a, the interfacial area per unit of volume. There are common methods for KLa determination such as dynamic gassing out or sulfite reaction.

OUR is defined as the quantity of oxygen consumed by cells per unit of volume and per unit of time. OUR is constant for resting cells.

OTR in the reactor is defined as the quantity of oxygen transferred from the gas to the liquid phase per unit of volume and per unit of time.

In order to evaluate KLa coefficient, we used a biological oxygen consumption method based on the OTR/OUR equilibrium.

OUR determination

In a 350-ml glass bioreactor filled with the reference solution and saturated with oxygen by air sparging, a suspension containing 3 g l−1 of resting cells was added, then, air supply was stopped, and DOC was recorded on-line. Temperature of the system was maintained at 30°C and vibration frequency at 30 Hz. The value of OUR was calculated as the slope of the curve representing DOC versus time.

Dynamic equilibrium OTR/OUR method (DEOM)

The DOEM is based on the equilibrium state between OTR and OUR in a bioreactor containing resting cells (characterized by a constant OUR).

The vessel was first filled to working volume (1.5 l) with the reference solution. The liquid phase was then saturated with oxygen by flushing air through the liquid surface and maintaining the bioreactor at a constant vibration frequency until DOC reached 100%.

The overpressure in the vessel was fixed at 10 mbar and the temperature controlled at 30°C. The bioreactor was inoculated, and DOC was measured then plotted against time into a curve.

Each vibration frequency was characterized by a new equilibrium state corresponding to a CL value.

KLa values were finally calculated from the equations below:
$${\text{OTR}} = K_{{\text{L}}} a{\left( {C^{ * }_{L} - C_{{\text{L}}} } \right)}$$
(1)
where OTR is the oxygen transfer rate; KLa is the gas–liquid interface transfer coefficient; C*L is the saturation concentration of oxygen in liquid phase and CL is the concentration of oxygen in liquid phase.
At equilibrium:
$$ {\text{OUR}} = {\text{OTR}} $$
(2)
Substituting OTR from Eq. (1), in Eq. (2) we get:
$${\text{OUR}} = K_{L} a{\left( {C^{ * }_{L} - C_{{\text{L}}} } \right)}$$
Then,
$$K_{L} a = {{\text{OUR}}} \mathord{\left/ {\vphantom {{{\text{OUR}}} {{\left( {C^{ * }_{L} - C_{{\text{L}}} } \right)}}}} \right. \kern-\nulldelimiterspace} {{\left( {C^{ * }_{L} - C_{{\text{L}}} } \right)}}$$
(3)

Thus, a KLa value was calculated for each vibration frequency of the system.

Viscosity and flow behaviour

To alter the viscosity of the reference solution, two polymers were tested: hydroxymethylcellulose (HMC) purchased from Clariant France and carboxymethylcellulose (CMC) purchased from Sigma Chemicals, USA.

CMC was progressively dissolved in distilled water at 40°C, then, 9 g/l of NaCl and 3 g/l of sucrose were added. A range of CMC concentrations was performed: 0, 3, 5, 7, 10 and 15 g/l. HMC was prepared under the same conditions.

The flow behaviour of the fluids tested was determined at 30°C in a rotational viscosimeter (Haake Rheostress®1, Germany) using the double concentric cylinders fixtures DG 41 (sample volume = 6.5 ml, radii ratio = 1.02, gap ext. = 0.35 mm, gap int. = 0.25 mm) in the shear rate range of 50–1,500 s−1.

Results

Pressure effect

Feasibility studies have first demonstrated that pressure inside the pouch was an important parameter. It affects both the contact surface between the flexible pouch and the vibrating table and the oxygen solubility in the liquid phase.

Thus, the effect of pressure on oxygen transfer inside the pouch was studied. The experiments were carried out at 30°C at a constant airflow of 0.15 l min−1 (0.1 vvm) at different vibration frequencies and different overpressure values. OUR of the cells and volumetric transfer coefficient (KLa) were measured as described previously in “Materials and methods”.

Figure 2 shows the effect of overpressure inside the pouch on the KLa at different vibration frequencies of the system. Oxygen transfer data were obtained in a 5-l bioreactor containing 1.5 l of liquid and 3 g l−1 of yeast resting cells.
https://static-content.springer.com/image/art%3A10.1007%2Fs00253-006-0682-1/MediaObjects/253_2006_682_Fig2_HTML.gif
Fig. 2

Volumetric transfer coefficient vs overpressure in the bag at 30°C, 0.1 l/min of aeration rate at different vibration frequencies

The overpressure in the bioreactor chamber was varied step by step from 0 to 20 mbar and the frequency from 20 to 30 Hz.

The maximum KLa measured was close to 85 h−1 corresponding to 10 mbar of overpressure. For all tested vibration frequencies (20, 25 and 30 Hz), the KLa increased when the pressure inside the chamber of the bioreactor increased until 10 mbar; over this value, the KLa dropped when the pressure continued to increase.

For this reason, it was decided to run all the other experiments at an overpressure of 10 mbar. Fixing a constant overpressure in the pouch is fundamental in this study in order to keep a constant contact surface between the pouch and the vibrating table.

Mixing time

Mixing time was measured by injecting a pH effector HCl (3 N) and recording on-line the variation of pH in the pouch until its stabilization (the experiments were duplicated using NaOH, 3 N). The experiments were carried out using the reference medium as liquid phase, at 30°C, 10 mbar and at different vibration frequencies (25, 30 and 35 Hz).

Mixing time (defined as time for complete homogeneity) ranged from 30 to 80 s for 2 l of liquid (in a 5-l pouch) depending on the vibration frequency of the table. Results are shown in Table 1. These mixing time values are acceptable for a bioreactor without mechanical agitation; in our case, the mixing was only induced by the wave propagation.
Table 1

The mixing time in the pouch at different vibration frequencies

Vibration frequency (Hz)

Mixing time (s)

25

52

30

39

35

16

Oxygen transfer coefficient in liquid phase (KLa)

Effect of vibration frequency

The KLa was measured as a function of the vibration frequency of the table. The experiment was carried out in a 5-l bioreactor containing 1.5 l of the reference solution and inoculated with 3 g/l of fresh yeast resting cells. Temperature was maintained at 30°C, overpressure at 10 mbar, and air feed rate was 0.15 l min−1 (0.1 vvm).

The vibration frequency was varied from 15 to 30 Hz. The resting cells oxygen demand was constant during the experiment, and its measured value was 64 mg g−1 h−1. Results are given in Fig. 3.
https://static-content.springer.com/image/art%3A10.1007%2Fs00253-006-0682-1/MediaObjects/253_2006_682_Fig3_HTML.gif
Fig. 3

Profile of dissolved oxygen vs time at different vibration frequency (15, 20, 25 and 30 Hz) and corresponding KLa at 30°C, 10 mbar of overpressure and 0.1 vvm of aeration rate

The KLa performed by the system ranged from 8 to 80 h−1. The oxygen transfer coefficient increased with the vibration frequency of the system. Aunins et al. (1989) reported a KLa of about 1 h−1 for 500-ml liquid in a 1,000-ml spinner; Dorresteijn et al. (1994) reported a value of around 2 h−1 for 300 ml of liquid in a 600-ml spinner system and Singh (1999) reported a KLa of 3.5 h−1 for 500 ml in a 2-l wave bioreactor system.

Effect of volume

The oxygen transfer coefficient, KLa, was measured at different volumes ranging from 1.5 to 4 l in a 5-l pouch. The experiment was performed under the same conditions as the previous one, at temperature of 30°C, pressure in the bioreactor of 10 mbar, 0.15 l/min of air feed and inoculated with 3 g/l of fresh yeast. The measured cell’s oxygen demand was 64 mg g−1 h−1, and the vibration frequency of the system was fixed at 25 Hz. Results are shown in Table 2.
Table 2

The oxygen transfer in a 5-l bioreactor at different working volumes

Volume (ml)

CL (%)

KLa (h−1)

Filling coefficient (%)

1,500

89

71.02

30

2,000

85.2

52.79

40

2,500

83.4

47.06

50

3,000

79.8

38.68

60

3,500

57.5

18.38

70

4,000

25

10.42

80

The results showed that while increasing the liquid volume in the bioreactor, the oxygen transfer decreased. This was due to the reduction of the gas/liquid transfer surface and the increase of the liquid depth. The transfer of oxygen from the surface to the bottom of the liquid is longer (no bubbling).

For a filling coefficient of 30% and a vibration frequency of 25 Hz, the maximum KLa measured was about 71.02 h−1, and the minimum KLa measured was around 10 h−1 for a filling coefficient of 80%.

Effect of viscosity

In a previous experiment not reported here, it was demonstrated that in an agitated vessel, both HMC and CMC have no affect on the OUR of the yeast resting cells.

Thus, the rheology of the medium was modified through the addition of these polymers without adversely affecting the cell’s activity. These chemicals have also been shown to be non-toxic, even to sensible cells such as mammalian cell types (Sen et al. 2002), and are able to influence the medium viscosity (Moreira et al. 1995). The study also demonstrated that the rheology of both HMC and CMC solutions is Newtonian.

But because HMC leads to high foaming from 5 g/l (13.5 cP) concentration and on, CMC was chosen as a test additive to study the effect of viscosity up to 46.5 cP (refer to Table 3).
Table 3

Effect of viscosity on the volumetric oxygen transfer at different vibration frequencies of the system

CMC concentration (g l−1); Viscosity (cP)

Vibration frequency (Hz)

KLa (h−1)

0; 1.025

15

8

20

11

25

26

30

80

15

7

1; 1.35

20

9.61

25

36.18

30

71.64

3; 3.26

20

9.78

25

22.67

30

68.33

5; 5.9

25

15.99

30

49.11

10; 18

25

28.65

30

7.36

15; 46.5

  

30

9.15

KLa was measured in the medium at different concentrations of CMC and at different vibration frequencies of the system. DO profile was recorded, and then, KLa was calculated using the DOEM (as described in “Materials and methods”). The temperature was maintained at 30°C, the culture volume was 1.5 l in a 5-l bioreactor, and the measured cell’s oxygen demand was around 55 mg g−1 h−1. The curves corresponding to a high and a low tested viscosity are shown in Fig. 4.
https://static-content.springer.com/image/art%3A10.1007%2Fs00253-006-0682-1/MediaObjects/253_2006_682_Fig4_HTML.gif
Fig. 4

Variation of dissolved oxygen profile and KLa at different vibration frequencies of the: at low (3.2 cP) and high (45 cP) viscosity

At a vibration frequency of 30 Hz, the oxygen transfer coefficient dropped from 68 h−1 at low viscosity (3.2 cP) to 9 h−1 at high viscosity (45 cP).

Discussion

According to Henry’s law usually defined as:
$$C^{ * } = P_{g} \cdot K_{{\text{H}}} $$

Here, C* is the maximum oxygen concentration in the aqueous phase, and Pg is the partial pressure of oxygen in the gas phase in contact with liquid.

The increase of pressure in the flexible pouch results in an increase of the maximal oxygen solubility (C*) in the liquid phase, which can affect the KLa of the system:
$$K_{{\text{L}}} a = {{\text{OUR}}} \mathord{\left/ {\vphantom {{{\text{OUR}}} {{\left( {C^{*}_{L} - C_{{\text{L}}} } \right)}}}} \right. \kern-\nulldelimiterspace} {{\left( {C^{*}_{L} - C_{{\text{L}}} } \right)}}$$

But due to the low pressure inside the pouch (0–20 mbar), the C* could be considered as constant.

Then, the recorded change in KLa as a function of overpressure is mainly due to the adhesion of the flexible pouch on the vibrating table and not to the low variation in oxygen solubility. The adhesion of the flexible bioreactor was shown to be optimal at 10 mbar of pressure.

When the pressure inside the bioreactor chamber exceeds 10 mbar, this causes the wave’s crushing at the liquid surface, which can be visually observed, and this reduces the liquid/gas transfer surface and then the KLa values. This phenomenon is now under investigation to be explained and quantified.

In addition, at a constant pressure, the increase of the vibration frequency of the system induces an increase in the KLa (see Fig. 3). These results have to be compared with those obtained with other similar surface aerated systems (see Table 4).
Table 4

KLa vibrating pouch versus wave bioreactor and spinner flasks (Modified from Singh 1999)

Liquid (ml)

KLa in the vibrating pouch ( h−1)

KLa in the wave bioreactor ( h−1)

KLa in the spinner ( h−1)

500

51

3.5

1.15

1,000

76

3

0.9

10,000

4

With 2.5 l of liquid in a 5-l bioreactor, the measured KLa was around 24 h−1; whereas, in a comparable wave bioreactor with 10 l of liquid in a 20-l bioreactor, the maximum KLa measured was around 4 h−1 (Singh 1999). It was reported also that in a 1-l spinner, the measured KLa was around 1 h−1, and in a 10-l one, the calculated KLa was less than 0.4 h−1, making it essentially useless for cell cultivation. After such comparison, it is obvious that the vibrating system is currently the most efficient system for surface oxygen transfer.

Therefore, if oxygen transfer was limited in these systems, then, higher cell density in the vibrating pouch could be obtained.

Unlike aerated stirred tank bioreactor, there is no bubble damage, no vortex or danger of “overaeration”, but shear effect on animal cells still has to be evaluated.

At high viscosity (47 cP), two phenomena were observed. First, the oxygen transfer coefficient decreased markedly with the increase of viscosity at constant vibration frequency of the system. This can be due to the reduction of the surface of oxygen transfer because of the increase of the medium consistency, as it was visually observed during the experience. Second, the signal of the dissolved oxygen probe was not stable (Fig. 4). This could be explained by the fact that the increase of viscosity increased the thickness of the stagnant film near the dissolved oxygen probe membrane, which made the renewal time of the interface between the liquid and the membrane longer.

Table 3 summarizes the effect of the liquid’s viscosity and the vibration frequency on the KLa of the system.

Conclusion

A novel culture system has been developed and characterized in this study; the used bioreactor is a flexible disposable plastic pouch equipped with special ports for sterile additions, sampling and air feeding. The bioreactor is fixed on the plate-form of a vibrating table; mixing and oxygen transfer are accomplished by the vibrating motion induced by the plate-form.

In order to characterize the system, the mixing time and the viscosity effect on KLa were evaluated. The oxygen transfer has been optimized in terms of vibration frequency and overpressure inside the pouch; a KLa of 80 h−1 was reached at 30 Hz and 10 mbar.

The obtained results demonstrated that depending on the critical oxygen concentration, OTR could reach up to 640 mg l−1 h−1, which is good enough to satisfy the OUR of animal and plant cells culture. The reactor is now being scaled up at 200 l for the culture of micro-aerophelic strains of bacteria and for hairy-roots.

Copyright information

© Springer-Verlag 2006