1.
52.2

9.....iF:

. ‘1 .4 5|
5

Li.

A . Erna
; I 300.

. I Irrdm

.11.?!“
if I : .
.- PILIJ,.J£.IHLMH:.

tan; IquW-vrﬁruﬂ. 4T.

Lib

. I?

. Arr A
J3?! ... 5..
$3.... ..
.J M... .1. «J...
. . 1.1mm e..w$va:.mm.¢n1u_
... , ’l I15 . I
.z A I L .
.uuiduﬁutuuuw .Bou i . y . u
bl: I. fl . .
. N‘ﬂﬂﬂmi. Inuit)! . w

o; '15.” -

ﬂy...

.4!!!-

 

.J.‘ .491} 3: .. .
3.1 .n . l. '0'

 

Hilary? . a

I’ .. 7.
‘ 151.: Lult...‘

‘ I.

1
7 ~95. .
{in

3351.0...

ms . IllllllllllllllllllHI!”llllllllllIHHIHlllillllllllllil

A, 1293 01581 2203

 

LIBRARY

Michigan State
Universlty

 

 

 

This is to certify that the

dissertation entitled

A Continuous Magnetofluidized Bed Bioreactor for
Production of Plant-Cell Secondary Metabolites

presented by

Tyler T. Ames

has been accepted towards fulﬁllment
of the requirements for

Ph . D . degree in Chemical Engineering

 

 

 

Major professor

Date May 1, 1997

MS U i: an Afﬁrmative Action/[q ual Opportunity Institution 0-12771

 

PLACE IN RETURN BOX to remove this checkout from your record.
TO AVOID FINES return on or before date due.

__ Ii l.._l
ll:_ll =:
__—| ﬂ
ll —|-L_J

:Iﬁl -
:L__l:]

MSU le An Affirmative Action/Equal Opportunity Institution

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

A CONTINUOUS MAGNETOFLUIDIZED BED BIOREACTOR FOR
PRODUCTION OF PLANT-CELL SECONDARY METABOLITES

By

Tyler T. Ames

A DISSERTATION
Submitted to
Michigan State University
in partial fulﬁllment of the requirements
for the degree of

DOCTOR OF PHILOSOPHY

Depanment of Chemical Engineering

1997

ABSTRACT

A CONTINUOUS MAGNETOFLUIDIZED BED BIOREACTOR FOR
PRODUCTION OF PLANT-CELL SECONDARY METABOLITES

By

Tyler T. Ames

Plant cells have the potential to produce a wide variety of economically important
products. The magnetoﬂuidized bed bioreactor (MF B) which offers continuous, plug-
ﬂow biocatalyst throughput, is well suited for production of intracellularly stored plant
metabolites. A MFB system, featuring a magnetic valve for solids (MVS), has been
assembled, operated, and reﬁned. The MFB has been operated with soybean cell culture,
producing daidzein and genistein for up to 2 months. Soybean cells were co-immobilized
with fertitic stainless steel in calcium alginatc beads. Biocatalyst microscopy and Thiele
modulus calculations indicated that oxygen diffusion in the biocatalyst was not rate
limiting in this system. Continuous operation of the MFB, featuring dissolved 02 and
C02 control, has been demonstrated past start-up to the point of steady growth rates. An
unsteady-state model of the system provided insight on experimental data as well as
protocol for steady-state operation. MFB feasibility was conﬁrmed by the scale-up
calculations. Power per volume requirements decrease with size and microbubble aeration
was demonstrated as an efficient method for providing oxygen at large scale. MFB
development was paralleled by development of the Diffusion Gradient Chamber (DGC)
for rapidly optimizing plant cell culture growth and product formation. The DOC

provides exposure of immobilized plant cells to a continuum of essentially steady-state

concentrations in a single experiment. Image acquisition and analysis provided a non-
invasive method for tracking growth. Growth and isoflavonoid accumulation responded to
several chemical effector gradients including carbon source, nitrogen source, basal salts,
hormones, antibiotics, and an. isoﬂavonoid precursor. Optimal conditions were generally

within normal ranges and did not provide a new medium formulation for the MFB.

ACKNOWLEDGMENTS

I would like to thank Dr. Mark Worden for his guidance, dedication, and support
throughout my doctoral program. I also thank Dr. Ken Sink for his collaboratory efforts
and of course for our soybean cultures. I greatly appreciate the advice and job search
assistance given by Dr. Bruce Dale. I would like to thank Dr. Daina Briedis not only for
participation on my guidance committee, but also for her role in my choice of MSU for
graduate school.

I would like to thank Mark Widman, Marshall Bredwell, and Mark Mikola for
countless instances of assistance and humor over the past three years in the lab. We have
had the privilege of several undergraduate researchers, all appreciated, but especially
Casey Preston for his dedicated work and good spirit. I would also like to thank Joseph
Leykam for graciously providing mass spectrometry method development and analysis.
Funding for this work was provided by the MSU Crop and Food Bioprocessing Center,
Center for Plant Lipids and Starches, and the MSU Biotechnology Training Program.

I am especially grateful to Julie Caywood, Candy McMaster, and Faith Peterson
for helping with innumerable hurdles. As always, I am forever grateful to my mother and

father for giving me this opportunity and believing in me.

iv

TABLE OF CONTENTS

LIST OF TABLES .............................................................................................................. ix
LIST OF FIGURES .............................................................................................................. x
1. INTRODUCTION .......................................................................................................... 1
2. OBJECTIVES AND SIGNIFICANCE ........................................................................... 3
2.1 Continuous MFB Operation ..................................................................................... 3
2.2 Bioreactor Scale Up ................................................................................................... 5
2.3 Diffusion Gradient Chamber for Plant Cell Culture .................................................. 6

3. BACKGROUND ............................................................................................................ 8
3.1 Fluidized Beds for Immobilized Plant Cells .............................................................. 8
3.2 Magnetoﬂuidized Bed (MF B) ................................................................................... 9
3.2.1 Principle of Operation ........................................................................................ 9
3.2.2 Features as a Bioreactor .................................................................................... 11

3.3 Microbubble Gas-Liquid Mass Transfer ................................................................. 13
3.4 Isoﬂavonoids ........................................................................................................... 15
3.4.1 Medical Applications ....................................................................................... 16
3.4.2 Agricultural Applications ................................................................................. 17

3.5 Soybean Culture ...................................................................................................... 18
3.5.1 Development .................................................................................................... 18
3.5.2 Enhancement of Isoﬂavonoid Synthesis ........................................................... 19

3.6 Diffusion Gradient Chamber ................................................................................... 22
3.6.1 Description ....................................................................................................... 22
3.6.2 Microbial Research ........................................................................................... 24

4. MFB DEVELOPMENT AND CONTINUOUS OPERATION ................................. 25
4.1 Introduction ............................................................................................................. 25
4.2 Materials and Methods ........................................................................................... 26
4.2.1 Chemicals .......................................................................................................... 26
4.2.2 Equipment ........................................................................................................ 26
4.2.2.1 MFB System ............................................................................................. 26
4.2.2.2 HPLC ......................................................................................................... 31
4.2.2.3 Gas Chromatography ................................................................................ 31

4.2.3 Plant Cell Culture and Analysis ....................................................................... 31

4.2.4 Respiration ....................................................................................................... 32
4.2.5 Soybean Culture ............................................................................................... 32
4.2.6 Immobilization .................................................................................................. 33
4.2.7 Cell Recovery ................................................................................................... 35
4.2.8 Analytical Procedures ....................................................................................... 35
4.2.8.1 Dry Weight ................................................................................................ 35
4.2.8.2 Biocatalyst Microscopy ............................................................................ 35
4.2.8.3 Extraction of Isoﬂavonoids ........................................................................ 36
4.2.8.4 Isoﬂavonoid Analysis ................................................................................ 36
4.2.8.5 Respiration ................................................................................................ 37
4.2.9 Batch Shake Flask Cultures .............................................................................. 37
4.2.10 MFB Operation .............................................................................................. 38
4.2.10.1 Start-up .................................................................................................... 38
4.2.10.2 Continuous Biocatalyst Processing ......................................................... 39
4.2.11 Theoretical ...................................................................................................... 41
4.2.11.1 Specific Growth Rates ............................................................................. 41
4.2.11.2 Thiele Modulus Calculations ................................................................... 42

4.3 Results and Discussion ............................................................................................ 43
4.3.1 Batch Shake Flask Cultures .............................................................................. 43
4.3.1.1 Immobilization ........................................................................................... 43
4.3.1.2 Antimicrobial Agents ................................................................................. 45
4.3.2 Asepsis Testing ................................................................................................ 46
4.3.3 Continuous Operation ...................................................................................... 47
4.3.3.1 Soybean Growth ........................................................................................ 47
4.3.3.2 Respiration and Oxygen Diffusion ............................................................ 49
4.3.3.3 Isoﬂavonoid Production ............................................................................ 52
4.3.4 System Improvements ...................................................................................... 54
4.4 Conclusions ............................................................................................................. 55
5. PROTOCOL AND MODELING FOR LONG-TERM MFB OPERATION ............ 56
5.1 Introduction ............................................................................................................. 56
5.2 Materials and Methods ........................................................................................... 57
5.2.1 Chemicals .......................................................................................................... 57
5.2.2 MFB System .................................................................................................... 57
5.2.3 Soybean Culture and Medium .......................................................................... 58
5.2.4 Batch Shake Flask Cultures .............................................................................. 58
5.2.5 Analytical ......................................................................................................... 60
5.2.6 MFB Operation ................................................................................................ 60
5.3 Model ...................................................................................................................... 61
5.3.1 Scope and General Assumptions ...................................................................... 61
5.3.2 Mathematical Development ............................................................................. 63

5.3.3 Solution Method ............................................................................................... 67

5.4 Results ..................................................................................................................... 67
5.4.1 Model Parameters ............................................................................................. 67
5.4.2 Biocatalyst Loading and Growth ...................................................................... 70
5.4.3 Sugar Hydrolysis and Consumption ................................................................ 72
5.4.4 Head Space Composition and Dissolved Oxygen ............................................ 73
5.4.5 Isoﬂavonoid Production ................................................................................... 75
5.4.6 Modeling of Steady State Operation ................................................................ 79

5.5 Discussion ............................................................................................................... 82

5.6 Conclusions ............................................................................................................. 86

5.7 Nomenclature ........................................................................................................... 87
5.7.1 Variables ........................................................................................................... 87
5.7.2 Subscripts ......................................................................................................... 88

6. MFB BIOREACTOR SCALE UP ............................................................................... 89

6.1 Introduction ............................................................................................................. 89

6.2 Materials and Methods ........................................................................................... 90

6.3 Theoretical ............................................................................................................... 91
6.3.1 Oxygen Depletion in the Liquid Phase ............................................................. 91
6.3.2 Solenoid Equations ........................................................................................... 92
6.3.3 Economic Analysis ........................................................................................... 95
6.3.4 Comparison with CSTR F ermentors ................................................................ 96

6.4 Results and Discussion ............................................................................................ 97
6.4.1 Oxygen Depletion in the Liquid Phase ............................................................. 97
6.4.2 Reduction in Power Consumption by Microbubbles ....................................... 97
6.4.3 Uniformity of Magnetic Field .......................................................................... 99
6.4.4 Power Per Volume Ratios ............................................................................... 100
6.4.5 Capital Costs .................................................................................................. 102
6.4.6 Power Beneﬁts of Microbubble Aeration ...................................................... 102

6.5 Conclusions ........................................................................................................... 103

7. PLANT STUDIES IN THE DGC .............................................................................. 105

7.1 Introduction ........................................................................................................... 105

7.2 Materials and Methods ......................................................................................... 106
7.2.1 Chemicals ........................................................................................................ 106
7.2.2 Diffusion Gradient Chamber System ............................................................. 106
7.2.3 Chromatography ............................................................................................. 109
7.2.4 Soybean Culture ............................................................................................. 109
7.2.5 Grayscale and Concentration Correlation Experiments .................................. 110
7.2.6 Soybean Growth Experiments ........................................................................ 110
7.2.7 Image Acquisition and Processing .................................................................. 111
7.2.8 Analytical ....................................................................................................... 112

Vii

7.2.8.1 Post-experiment Sampling of Gel ............................................................ 112

7.2.8.2 Sugar Assays ........................................................................................... 112
7.2.8.3 Ampicillin Assays ................................................................................... 113
7.2.8.4 Dry Weight Assays of Gel Sections ........................................................ 113
7.2.8.5 Isoﬂavonoid Assays of Gel Sections ....................................................... 114

7.3 Theoretical ............................................................................................................. 114
7.3.1 Unsteady-State Model of the DGC System .................................................. 114
7.3.2 Parameter Determination ................................................................................ 116
7.4 Results ................................................................................................................... 117
7.4.1 Oxygen Limitations in the Immobilization Layer ........................................... 117
7.4.2 Simulation of Unsteady-State Diffusion Gradients ........................................ 118
7.4.3 Grayscale and Concentration Correlation ....................................................... 119
7.4.4 Growth Studies ............................................................................................... 120
7.4.4.1 Linearity of Diffusion Gradients ............................................................. 120
7.4.4.2 Effect of Sucrose ...................................................................................... 121
7.4.4.3 Effect of l-B5 Medium Components ...................................................... 125
7.4.4.4 Effect of Antibiotics ................................................................................ 127
7.4.5 Isoﬂavonoid Production Studies ..................................................................... 129
7.5 Discussion ............................................................................................................. 134
7.5.1 Correlation of Grayscale and Concentration .................................................. 134
7.5.2 Suitability of DGC for Growth Studies ......................................................... 135
7.5.3 Non-Ideality of DGC Reservoirs ................................................................... 136
7.5.4 Determination of Optimum Growth Conditions ............................................ 136
7.5.5 Determination of Optimum Isoﬂavonoid Conditions .................................... 137
7.6 Conclusions ........................................................................................................... 139
8. SUMMARY AND CONCLUSIONS ......................................................................... 139
9. LIST OF REFERENCES ............................................................................................. 142

viii

LIST OF TABLES

Table 1. Volumetric productivity calculation for MFB. ..................................................... 4
Table 2. Composition of soybean medium ........................................................................ 33
Table 3. Antibiotics successfully tested with soybean cell culture ................................... 46
Table 4. Non-adjusted model parameters. ......................................................................... 69
Table 5. Adjusted model parameters. ................................................................................ 70
Table 6. Parameters for present value cost analysis .......................................................... 95
Table 7. List of chemical gradients used in DGC experiments. ....................................... 111

LIST OF FIGURES

Figure 1. Magnetic ﬁeld lines and elimination of mixing in a MFB. .................................. 10
Figure 2. Plug-ﬂow movement of solids through a MFB. ................................................. 11

Figure 3. Hypothetical plant-cell growth and secondary-product-formation kinetics. ..... 13

Figure 4. Structure of the isoﬂavonoids daidzein and genistein ......................................... 16
Figure 5. Diffusion Gradient Chamber. ............................................................................. 23
Figure 6. Diagram of the magnetoﬂuidized bed bioreactor. ............................................... 27
Figure 7. Cross-sectional view of MFB head plate. .......................................................... 28
Figure 8. Cross-sectional view of MFB base plate. .......................................................... 29
Figure 9. 3-D diagram of the magnetic valve for solids ...................................................... 30
Figure 10. Beadmaker for co-immobilization of soybean and stainless steel. ................... 34
Figure 11. Immobilized soybean cell densities for replicate shake ﬂasks .......................... 44
Figure 12. Product biocatalyst loading as dry weight per biocatalyst volume. ................. 47
Figure 13. Soybean speciﬁc growth rates in the MFB. ..................................................... 50
Figure 14. Product biocatalyst cross-sections at 40 times magniﬁcation. ......................... 52
Figure 15. Isoflavonoid content of product soybean as mass per dry tissue. ................... 53
Figure 16. Diagram of the magnetoﬂuidized bed bioreactor system .................................. 59
Figure 17. Biocatalyst plant-cell loadings and speciﬁc grth rates ................................. 71

Figure 18. Data and model predictions for plant-cell loading of each biocatalyst plug. 72

Figure 19. Sucrose, glucose, and fructose concentrations in MF B medium. ..................... 74
Figure 20. Oz and C02 content of head space, and dissolved 02 (DO) ............................. 75
Figure 21. Isoﬂavonoid concentrations in product biocatalyst soybean. .......................... 76

X

Figure 22. Effect of antibiotic cocktail on soybean cell growth ......................................... 77

Figure 23. Effect of antibiotic cocktail on daidzein concentrations in soybean ................. 78
Figure 24. Effect of antibiotic cocktail on genistein concentrations in soybean. ............... 78
Figure 25. Total volume of biocatalyst and volumetric productivity ................................ 79
Figure 26. Extended simulation of MFB run. .................................................................... 81
Figure 27. Simulation of MFB using a K", of 0.5 g/L and 20 g/L sucrose feed. ................. 81
Figure 28. Simulations of MFB for 9 and 12 day biocatalyst residence times .................. 82
Figure 29. Comparison of saturation mode] with and without inhibition. ........................ 85
Figure 30. Power consumption required to prevent mixing at top view port. .................. 98
Figure 31. Power consumption required to prevent mixing at bottom view port. ............ 98
Figure 32. Field strength as a function of column height for an aspect ratio of 2 .............. 99
Figure 33. Field strength as a function of column height for an aspect ratio of 8 ............ 100
Figure 34. Power consumption of 100 Gauss MFB vs. CSTR ....................................... 101
Figure 35. Power consumption of 300 Gauss MF B vs. CSTR ....................................... 101
Figure 36. Diagram of the Diffusion Gradient Chamber top view. ................................. 107
Figure 37. Schematic side-view of DGC showing three slabs. ........................................ 108
Figure 38. Simulation of approach to steady-state .......................................................... 118
Figure 39. Correlation of image grayscale with plant-cell density ................................... 1 19
Figure 40. Total sugar gradient data and model for a 0 to 20 g/L experiment. ................. 120
Figure 41. Total sugar gradient data and model for a 0 to 30 g/L experiment. ................. 121
Figure 42. Image from day 13 of a 0 to 20 g/L sucrose experiment. ................................ 122
Figure 43. Grayscale proﬁles for a 0 to 20 g/L experiment. ............................................ 122
Figure 44. Grayscale proﬁle from day 14 of a 0 to 30 g/L experiment ............................ 123
Figure 45. Image from day 14 of a 0 to 30 g/L sucrose experiment. ................................ 124

Figure 46. Grayscale proﬁles for inlet and outlet halves of the DGC. ............................ 124

Figure 47. Total sugar assays for inlet and outlet halves of the DGC ............................. 125
Figure 48. Image from day 13 of Gamborg basal salts experiment. ................................. 126
Figure 49. Grayscale proﬁle from day 13 of Gamborg basal salts experiment ................ 126
Figure 50. Grayscale proﬁle for day 11 of a Timentin experiment. ................................ 127

Figure 51. Grayscale proﬁles for day 15 of replicate 300 ppm ampicillin experiments. 128

Figure 52. Grayscale proﬁle for day 23 of 600 ppm ampicillin experiment. .................. 129
Figure 53. Free isoﬂavonoid concentrations for 0 to 1x nitrogen gradient ....................... 130
Figure 54. Free isoﬂavonoid concentrations for 0 to 3x nitrogen gradient ....................... 130
Figure 55. Total isoﬂavonoid concentrations for 0 to 5x nitrogen gradient. .................... 131
Figure 56. Free isoﬂavonoid concentrations for 0 to 1x thiamine gradient. ..................... 132
Figure 57. Total isoﬂavonoid concentrations for 0 to 5x 2,4-D gradient ......................... 132

Figure 58. Free isoﬂavonoid concentrations for 0 to 1 g/L phenylalanine gradient. ........ 133

Figure 59. Total isoﬂavonoid concentrations for 0 to 0.1 g/L phenylalanine gradient. 133

l . INTRODUCTION

Plant cells have the potential to produce a wide variety of economically important
products. However, commercial development of this potential is limited by slow growth
rates, relatively low rates of product formation, and consequently, low bioreactor
productivities (Sahai and Knuth, 1985). The research presented here addresses this
problem on two levels. First, a continuous bioreactor for plant-cell culture which offers
several advantages over traditional systems such as CSTRs was developed and
demonstrated. Second, a novel tool was developed for determining concentrations of
medium components that optimize growth and product expression. Soybean cell culture,
which produces the isoﬂavonoids genistein and daidzein, was used a model system. The
medical and agricultural potential of these compounds gives added value to the work,
which is relevant to a broad range of plant cell systems and products.

The research is presented in 6 sections. Section 2, Objectives and Signiﬁcance,
presents the goals and hypotheses of the research as well as the signiﬁcance of the work.
Background information and a review of relevant literature comprises Section 3,
Background Preliminary development of the magnetoﬂuidized bed bioreactor (MFB) is
described in Section 4, MP B Development and Continuous Operation. Section 5,
Protocol and Modeling for Long-Term Operation, expands upon key aspects of
operation, examines steady-state operation, and provides a mathematical description of
the system. Scale-up feasibility issues including use of a third (gas) phase and magnetic

ﬁeld generation are addressed in Section 6, MFB Bioreactor Scale Up. Development of

the Diffusion Gradient Chamber (DGC) as a tool for plant bioprocess is detailed in

Section 7, Plant Studies in the DGC.

2. OBJECTIVES AND SIGNIFICANCE

2.1 Continuous MFB Operation

The primary objective of the research was to demonstrate continuous operation of
a MFB for production of isoﬂavonoids from soybean cell culture. Furthermore,
achievement of steady-state growth and product formation was a goal. Demonstration of
continuous, steady-state operation at the lab scale is considered an important prerequisite
for further development (scale-up) of the system. It was hypothesized that by
maintaining key process variables at steady values, growth and product concentrations
can be maintained at a steady-state. Key process variables were considered to be sucrose
and antibiotic concentrations, pH, temperature, dissolved oxygen, and carbon dioxide
content of the gas phase.

The MP B bioreactor is ideally suited for production of intracellular plant-cell
secondary metabolites because it provides continuous processing of immobilized cells
while maintaining a uniform residence time. By eliminating solids mixing the MFB has the
potential for a signiﬁcant increase in productivity over conventional ﬂuidized beds. Two
speciﬁc isoﬂavonoids produced by soybeans, genistein and daidzein, were the target
compounds for this research project. In contrast to the current extraction method for
isoﬂavonoids, the proposed bioprocess would allow soybean cell cultures to be
maintained under optimal conditions for isoﬂavonoid production, and would allow
continuous, rather than seasonal production by ﬁeld crops. Genistein and daidzein

concentrations in soybean cell culture (Zacharius and Kalan, 1990) have been reported as

one to two orders of magnitude higher than in soybeans (Wang and Murphy, 1994; Naim
et al., 1974). It is therefore anticipated that the proposed process could reduce the costs
of genistein and daidzein by more than an order of magnitude.

Shuler et al. (1990) has noted that the primary determinant of economic viability
for a bioprocess, the volumetric productivity of the reactor times the wholesale price of
the product, is 0.12 $/L-day +/- 60% for a wide variety of fermentations. A volumetric
productivity can be calculated for MF B productivity as shown in Table 1.

Table 1. Volumetric productivity calculation for MF B.

 

 

Variable Value Source
Value of Isoflavonoid $1000 per Sigma: $4,800 /g natural genistein
$700 /g synthetrc genistein

ICN: $1,400 /g daidzein

Concentration in soybean 600 ppm Genistein content in Worden lab
Glycine max

Reported as high as 15,600 ppm
(Zacharius and Kalan, 1990)

Soybean density in biocatalyst 10 g/L Achieved in MF B (Fall 1995)

Biocatalyst packing 0.7 L per L estimate

Biocatalyst residence time 20 days Used in MFB (Fall 1995), approx. end
of grth in batch cultures

Volumetric productivity 0.21 $/L-day

 

 

The calculated volumetric productivity falls within the range of commercial fermentations.

This result is encouraging as the values in Table 1 should be obtainable in a successful

steady-state run. Furthermore, these numbers do not take into account improvements

which may result ﬁ'om proposed research with the DGC.

2.2 Bioreactor Scale Up

For lab scale bioreactors using a continuous liquid recycle, pre-saturation of the
liquid phase with gaseous substrates may be adequate. Bramble et a1. (1990) conducted 2-
phase MFB experiments with C. arabica, using a dialysis cartridge to aerate the liquid
phase before it entered the reactor. However, for commercial-scale bioreactors where
higher cell densities and longer liquid residence times would be needed, this approach is
not feasible because the dissolved oxygen would be depleted from the liquid phase as it
rises through the bioreactor. Addition of gas is needed to replenish what is consumed by
the cells.

A secondary objective of the research was to demonstrate a useful pairing of
microbubble aeration and the magnetofluidized bed bioreactor. It was hypothesized that
power required to stabilize the bed will be reduced by using microbubbles instead of
conventional gas sparging. A positive result would translate to increased gas mass transfer
and reduced operating costs. Microbubble aeration is attractive method of supplying
oxygen for two reasons. First, microbubbles provide very high mass-transfer coefﬁcients.
Second, microbubble aeration causes less disruption of a magnetoﬂuidized bed than
traditional gas sparging.

The feasibility of a commercial scale MFB also depends on the economics scaling

up to large scale. Both power requirements for magnetic ﬁeld operation and capital costs

of construction are important issues. It was hypothesized that the MFB would scale
favorably by the criteria of decreasing power-per-volume ratio with increasing reactor
volume. It was further hypothesized that capital costs would be a- small fraction of
typical fermentation system capital costs. Validation of this hypothesis is signiﬁcant in

determining the commercial potential of the process.

2.3 Diffusion Gradient Chamber for Plant Cell Culture

An additional objective of the research was to develop the Diffusion Gradient
Chamber (DGC) as a tool for optimizing conditions in bioprocesses which utilize plant
cell culture. The DGC was explored as a highly efficient tool for collecting data on plant
cell response to environmental conditions. The stimulus which causes the response is
deﬁned as a chemoeffector. The DGC can be used to optimize concentrations of both a
positive chemoeffector (e. g. substrate such as sucrose) and a negative chemoeffector (e.g.
antibiotics inhibiting growth). Optimization is used here in a broad sense, not limited to
maximization. Optimization could include determination of the lowest sugar concentration
which yields growth at 95% of the maximum rate. Balancing antibiotic effectiveness
against microbes and plant toxicity (both dependent on concentration) is another example.
It was hypothesized that plant cell concentration (and thus growth) could be quantiﬁed
noninvasively with video capture and image analysis. The logical step beyond growth
optimization is optimization of product accumulation. It was hypothesized that
accumulation of genistein and daidzein could be optimized as a function of chemoeffectors

which may be important to secondary metabolism (e. g. phenylalanine) using the DGC.

- 1
511m

5‘ a
$1233.

. 1,1
C1119

The DGC allows plant cells to be grown under a spectrum of conditions
simultaneously. Image analysis and/or chemical assays are then used to determine the
effects of those conditions on cell growth and product formation. One DGC experiment
could potentially supply information equivalent to dozens of individual experiments in
Petri dishes or Erlenmeyer ﬂasks. This approach is expected to signiﬁcantly reduce the
time and effort required to determine environmental conditions which maximize growth
and/or product accumulation. Furthermore, success with product correlation has even
broader applications. Even if the product correlation hypothesis cannot be validated,
progress towards the objective may establish protocols for other plant cell systems where
more sensitive or non-invasive assays are available or where products are found in higher
concentrations. Examples could include: production of pigments which can be quantiﬁed
by color image analysis, production of valuable oils comprising a signiﬁcant fraction of
cell mass, or production of secondary metabolites for which a non-invasive assay
technique is available such as ﬂuorescence. The DGC provides an additional advantage
when applied to immobilized cell systems. As cells can be immobilized in the DGC gel,
information speciﬁc to a solid phase can be gained. Such information could include data on

cell mass and substrate gradients as well as effects of immobilization.

“for .
my. a.

5511

3. BACKGROUND

3.1 Fluidized Beds for Immobilized Plant Cells

Several reactor schemes are available for plant cell cultures, including stirred tanks,
bubble columns, air-lift reactors, hollow-ﬁber membranes, liquid-dispersed trickle and
incline reactors, and mist bioreactors for hairy root culture (McKelvey et al., 1993;
Whitney, 1992; Taya et al., 1989). However, each of these reactors has limitations, such
as cell death ﬁom shear, insufﬁcient mixing at high cell densities, dead zones, settling of
cells, and difﬁculty in scale-up (Panda et al., 1989). Furthermore, many of these reactors
are poorly suited for continuous operation in which the product is accumulated
intracellularly.

Fluidized bed bioreactors offer several advantages for immobilized plant cells.
Low shear forces and pressure drops prevent damage to ﬁ'agile cells. Excellent contact
between the gas, liquid, and solid phases eliminates the need for mechanical mixing. Shuler
et al. (1990) noted that high cell densities and continuous operation lead to increased
volumetric productivity over non-immobilized or batch processes. Cell densities may
reach 110 g/L in alginate beads, while suspension culture is limited to about 30 g/L (Sahai
and Knuth, 1985). The close proximity of immobilized cells allows intercellular
communication and transport that can be important for differentiation and secondary-
metabolite production (Shuler et al., 1990). Attachment and plugging caused by free cells

are eliminated by immobilization (Sahai and Knuth, 1985).

An important disadvantage of conventional immobilized systems, however, is that
most plant secondary metabolites are typically stored intracellularly. Soybean
isoﬂavonoids exist predominantly as glucoside-malonylated conjugates (Graham et a1 .,
1990; Kudou et al., 1991) which are located in the vacuole (Barz and Welle, 1992).
Permeabilization methods for release of products into the media result in low culture
viability. Brodelius (1990) notes that it appears very difﬁcult, if not impossible, to
release vacuolar substances into the medium by permeabilization of the plasma membrane
and the tonoplast without killing the cells.

Continuous solids throughput eliminates the need for metabolite release. Fresh
cells can be continually added at the reactor, maintained in the reactor for the optimal
length of time, and then removed for product recovery. The solids residence time can be
chosen to optimize the intracellular product concentration, and controlled by the rate of
solids throughput. Continuous operation also minimizes bioreactor down time, and thus

offers increased productivity over batch operation.

3.2 Magnetofluidized Bed (MFB)

3.2.1 Principle of Operation

A magnetoﬂuidized bed (MFB) consists of a ﬂuidized bed of magnetically
susceptible particles to which a direct-current (DC) magnetic ﬁeld is applied. The
magnetic ﬁeld induces dipoles in the particles and causes them to align along the ﬁeld lines
(Figure 1), thus eliminating the solids mixing that occurs in ﬂuidized beds. However, like
conventional ﬂuidized beds, the magnetized solids can still be passed though the bed. As a

result, illustrated in Figure 2, continuous solids throughput without mixing (i.e. "plug

13' 1

tier

if}.

F171

10

ﬂow") is possible (Rosensweig, 1978; Rosensweig, 1979; Rosensweig et al., 1981; Siegell,
1988; and Liu et al., 1991). Several applications have been found for MFB, and reviewed
by Liu et al. (1991). Magnetic ﬁelds have been successfully applied to ﬂuidized bed
bioreactors including continuous ethanol fermentation Meng et al., 1992), as well as
phenol degradation (Hu and Wu, 1987). Bramble et al. (1990) operated a two-phase
magnetically stabilized ﬂuidized bed with plant cells (C. arabica). However, none of

these bioreactors were operated in a solids-throughput mode.

Solenoid ' ‘ ’ 3" Magnetic
Field

Lmes

I

   

 

~ -oo-ooc-o- ’
---o-ooo-oo---
, "’.'.'.'.‘.“O*
I
‘ 2.31 . ;._n~,.;z.-.-.:...:.-c-.-1~:-:-: I-??:f'»:7:3’»:-:-T 742:3:71?;‘rf:?:?r-.f:?'f:.z.: :-::‘/_:?.7 -

 

‘\ Fluidized
- Bed

\

Figure 1. Magnetic ﬁeld lines and elimination of mixing in a MFB.

Preliminary MF B development has been ongoing in the research lab of Dr. Mark
Worden. Phase hold-up, liquid-mixing, and gas-mass-transfer properties of a prototype
MFB have been characterized (F reund and Worden, 1990; Thompson and Worden, 1991 ,
1992; Thompson, 1993; Thompson and Worden, 1997). These studies have indicated that
magnetization increases the gas volume fraction, decreases the axial liquid mixing, and

increases the volumetric gas-to-liquid mass-transfer coefﬁcient(k1a) in the ﬂuidized bed.

11

Particle Addition

ill

 

Particle Removal

Time ———>

Figure 2. Plug-ﬂow movement of solids through a MFB.

Magnetic valves for solids (MVS) have been developed to control movement of
solids out of ﬂuidized beds. An MVS consists of a magnetic-valve coil surrounding the
column in the vicinity of a magnetically susceptible screen. The screen openings are large
enough to allow free passage of solids when no current is applied to the valve coil.
Supplying power to the valve coil results in an induced magnetic ﬁeld at the screen which
stops solids ﬂow. Magnetic valves can be operated fully open, fully closed, or partially
open, to control solids ﬂow (Jaraiz-M. et al., 1984b). The MVS have no moving parts
and consume little power (Yang et al., 1982). Complete theory and design considerations
regarding screen sizes, capture and release currents, solids mass ﬂux, and magnetic coils
are given by Jaraiz-M. et al., (1984).

3.2.2 Features as a Bioreactor

The primary advantage of the MFB bioreactor is that it allows continuous

processing of the biocatalyst particles with tight control of their residence time in the

12

reactor. This capability, which has not been demonstrated for any other type of
bioreactor, allows the bioreactor productivity to be optimized, as illustrated in the
following example calculations.

The productivity of a bioreactor producing intracellular products is a function of
both the cellular kinetics and the residence-time distribution (RTD) of the biocatalyst.
The RTD, in turn, is determined by the degree of solids mixing as the solids pass through
the reactor. The mixing can be quantiﬁed either in terms of a Peclet number or a number of
stirred-tanks in series (F ogler, 1986). For a conventional ﬂuidized bed, assuming a four-
week growth and product-formation cycle for the plant cells, a typical dispersion
coefﬁcient of 4 cmZ/s (Dorgelo et al., 1985), and a bioreactor length of 60 cm, the number
of stirred tanks in series needed to approximate the solids mixing was calculated to be
1.000. Thus, the biocatalyst would become perfectly mixed, and the RTD would be
described by a stirred-tank model. In contrast, solids mixing is virtually eliminated in the
MFB, and the RTD would be approximated by a plug-ﬂow model. These two limiting
cases were used to demonstrate the increase in productivity possible with the MFB.
Figure 3 includes a lag phase, exponential growth, a stationary phase during which the
product is formed, and an exponential death phase during which the product decays
according to ﬁrst-order kinetics. Decrease in product concentration with time has been
observed for a variety of plant systems, including genistein and daidzein production by
Glycine max (Zacharius and Kalan, 1990). For these kinetics and a 4.8 week mean
residence time, a plug ﬂow bioreactor gives an average product concentration (and hence

bioreactor productivity) almost three times higher than a stirred-tank bioreactor. While

13

the actual increase in productivity realized by using the MFB will depend on the soybean
kinetics and the small degree of mixing at the ends of the column, this calculation indicates
the magnitude of improvement which may be expected.

A potential disadvantage of the MF B, which arises from its ability to
continuously process biocatalyst in plug ﬂow, is that it is impossible to change (program)
the medium composition in accordance with the age of the cells. Medium programming is

not needed for soybean isoﬂavonoid synthesis (Zacharius and Kalan, 1990) which is the

 

 

 

subject of this research.
1201- ..10
-- 9
100 --
Product " 3
3 so -1- " 7 Q
\ -- 6 £2
8 ..
g 60 " "' 5 é
a: - 4
0 4o . 2
Cells 3 IL
20 . " 2
.1. 1
0 J. i i l i i 0

 

0 1 2 3 4 5 6 7 8 910
Residence Time (weeks)

Figure 3. Hypothetical plant-cell growth and secondary-product—formation kinetics.

3.3 Microbubble Gas-Liquid Mass Transfer

Oxygen is commonly the rate-limiting nutrient in aerobic bioprocesses due to its

low aqueous solubility (about 8 mg/L for water in equilibrium air at 30°C and 1 atm

14

pressure). The main resistance to oxygen mass-transfer may lie in the gas-liquid interface,
the liquid-solid interface, or within the solid phase. For commercial-scale bioreactors, the
gas-liquid resistance is usually large, resulting in relatively low dissolved-oxygen
concentrations (van't Riet, 1983). In such cases, increasing the volumetric mass-transfer
coemcient (kla) increases the oxygen mass-transfer rate, and hence the reactor
productivity. An increase in k.a is typically achieved by either increasing the gas ﬂow rate
or the power-to-volume ratio (van't Riet, 1983). Both of these approaches work by
increasing the interfacial area between the gas and liquid. Unfortunately, they both add
signiﬁcantly to the operating costs of the process.

Microbubbles are surfactant-stabilized bubbles about 50 pm in diameter. The
surfactant layer surrounding microbubbles generates a diffuse electric double layer that
acts to repel other bubbles and prevent coalescence (Sebba, 1987). A microbubble
dispersion exhibits colloidal properties and is stable enough to be pumped (Longe, 1989).
The microbubble dispersion can thus be made remotely and then pumped, as needed, into
the bioreactor The effects of surfactant type and concentration, as well as salts, on
microbubble formation, drainage and stability have been characterized in the lab of Mark
Worden (Bredwell et al., 1995).

Conventional gas bubbles are about 100 times larger in diameter than microbubbles
(Kaster et al., 1990). Because surface area is inversely proportional to the bubble diameter
squared, microbubbles could potentially increase lqa values by orders of magnitude.
Signiﬁcant mass-transfer enhancement by microbubbles has been demonstrated in yeast

fermentations (Kaster et al., 1990). Microbubble dispersions are easily produced, are

15

biocompatible, and can produce extremely high k.a values, even in the absence of
mechanical agitation (Bredwell et al., 1995a; Bredwell et al., 1995b, Kaster et al., 1990).
Values for ha of up to 1800 h'1 in an unagitated bubble column using a gas volume
fraction less than 5% have been measured (Bredwell et al., 1995b). Furthermore, because
of their extremely small size, microbubbles equilibrate with the surrounding liquid phase
within about 15 s (Bredwell et al., 1994).

Power requirements for microbubble aeration are expected to be quite low during
scale-up. The Power number for our microbubble generator was measured as 0.036
(Bredwell et al., 1995b). Using this value, a speciﬁc oxygen-uptake rate of 10 mg 02/g
cells h, and previously reported details of microbubble formation (Bredwell et al., 1995a),
the power required to make enough microbubbles to sustain a 1 m3 reactor containing 30
g/L soybean culture was calculated to be 0.013 kW (i.e., 0.013 kW/m3) (Bredwell et al.,
1995b). By comparison, power-consumption rates for mechanically agitated microbial

fermentations are on the order of 1 kW/m3 (van’t Riet, 1983).

3.4 Isoflavonoids

Isoﬂavonoids constitute an important class of 15-carbon phenolic compounds
sharing a similar carbon skeleton. Found in all plant tissues, isoﬂavonoids have been
shown to exhibit a wide range of biochemical activities, including providing defense
against invading organisms (Rao, 1990). This research focused on the production of two

key isoﬂavonoids, genistein and daidzein, from soybean cell culture. The chemical

16‘

structures of these compounds are shown in Figure 4. These compounds, which are found

in high concentrations in soybeans, have important medical and agricultural applications.

DZidzein OH OH G:niste1n OH
Figure 4. Structure of the isoﬂavonoids daidzein and genistein.

3.4.1 Medical Applications

Recent studies have shown that the isoﬂavonoid genistein is a potent
chemotherapeutic agent that blocks several key steps in cancer induction and
proliferation. Protein tyrosine kinases (PTKs) play a key role in cancer induction,
proliferation, and metastasis (Akiyama et al., 1987; Mueller et al., 1992; Mustelin and
Burn, 1993). Genistein, a potent inhibitor of PTKs, has been shown to have antiproliﬁc
effects on human hepatocarcinoma (Mousavi and Adlercreutz, 1993) and stomach cancer
cells (Y anagihara et al., 1993). It can induce Erythroleukemia cells to differentiate into
non-cancerous Erythroid cells (Watanabe et al., 1991). Daidzein was also found to have
potent differentiation-inducing activity for human leukemia cells, both in vitro and in vi v0
(ling et al., 1993). Genistein has also been shown to block the invasion of normal cells by
tumor cells (Mueller et al., 1992) and block bacterial uptake by epithelial cells
(Rosenshine et al., 1992).

Topoisomerase is involved in breaking DNA for replication, transcription,

recombination, integration, all necessary steps in cell proliferation. Genistein limits the

l7

ability of ras oncogenes to transform cells by inhibiting topoisomerase, and this inhibition
was found to occur only in transformed cells (Okura et al., 1988). Angiogenesis
(formation of new capillaries) is necessary for proliferation and metastasis of tumors
beyond the 1-2 mm size. Genistein was found to inhibit angiogenesis of endothelial cells
in vitro as well as proliferation of several types of cancer (neuroblastoma,
rhabdyosarcoma, and Ewings Sarcoma) (Fotsis et al., 1993).

Evidence suggests that reactive oxygen species, such as hydrogen peroxide, play
an important role in mutagenesis and carcinogenesis, particularly in tumor promotion.
Reactive oxygen species may damage macromolecules such as proteins, RNA, and DNA,
which is related to carcinogenic initiation, promotion, and progression. Genistein has been
shown to inhibit both hydrogen peroxide production by leukemic cells and mouse skin
damage by hydrogen peroxide (Wei et al., 1993).

Isoﬂavonoids in soybean (primarily genistein and daidzein) may suppress
endogenous oestrogenic activity and inhibit hormone-dependent carcinogenesis of the
breast. A soybean rich diet has been associated with a reduced risk of breast cancer (Lee et
al., 1991). Rats showed a signiﬁcant decrease in chemically induced mammary cancers
when fed a soy diet (Messina and Barnes, 1991).

3.4.2 Agricultural Applications

Isoﬂavonoids such as genistein and daidzein are potential environmentally
compatible alternatives to synthetic pesticides for defense against pathogenic organisms.
Several studies suggest that isoﬂavonoids are part of the natural defense mechanisms of

plants. Soybean plants respond to infection or wounding by the production of the

18

enzymes phenylalanine lyase and chalcone-ﬂavanone isomerase, which are part of the
isoﬂavonoid synthesis pathway (Partridge and Keen, 1977). Genistein was found to
strongly inhibit the growth of Sclerotium rolfsii, which is responsible for southem-blight
disease (Gazaway and Hagan, 1989). Genistein has been shown to be toxic to
Phytophthora megasperma f. sp. glycina (Graham and Graham, 1991) which causes root
and stem rot in soybeans. In tissue infected by P. megasperma, production of
phytoalexins increased and there was rapid hydrolysis of isoﬂavone conjugates to
produce inhibitory levels of genistein at the site of infection (Graham and Graham, 1991).
Soybean isoﬂavones at 50 ppm of the growth medium inhibited growth of six species of
fungi by 60 to 85% (Naim et al., 1974). Genistein and daidzein concentrations of 50-75
ppm completely inhibited spore formation of Aspergillus niger, the causal factor in
deterioration of seeds in the ﬁeld following weather-delayed harvests (J acobsen and
Meronuck, 1989; and Adesanya et al., 1986).

The symbiotic relationship between soybean and Braafyrhizobium japonicum is
responsible for the formation of nitrogen-ﬁxing root nodules. Genistein and daidzein have
been shown to induce the nod genes of B. japanicum (Kosslak et al., 1987). Roots and
seeds exude these isoﬂavonoid compounds, suggesting that the plant attracts the nodule-

producing bacteria (Rao, 1990).

3.5 Soybean Culture
3.5.1 Development
Soybean tissue is rich in daidzein and genistein, the isoﬂavonoids targeted in this

proposal (Barz and Welle, 1992). Both compounds have been identiﬁed in hydrolyzed

l9

soybean root and leaf extracts (Porter et al. 1986), soybean seedling tissue and root
exudates (Graham, 1991b), and soybean roots (Cho and Harper, 1991). Genistein and
daidzein have also been found in soybean cell cultures in concentrations as high as 15,600
and 4410 pg per gram of dry cells, respectively (Zacharius and Kalan, 1990). Genistein
and daidzein in soybean are found together with their 7-O-glucoside 6"-O-malonate
conjugates (Barz and Welle, 1992). The conjugates can be hydrolyzed to ﬂee genistein

and daidzein by strong acid or almond B-glucosidase (Graham et al., 1990) for product

recovery.

Suspension cultures of Glycine max cv. Mandarin were provided by Dr. Ken Sink,
Department of Horticulture. Callus cells were initiated according to Fett and Zacharius
(1982). Brieﬂy, soybean seeds are surface sterilized and epicotyl, cotyledon or hypocotyl
explants are placed aseptically on B5 agar medium (Gamborg et al., 1968) with 1.0 or 2.0
mg/L 2,4-D. Calluses are subsequently transferred to B5 agar medium with 2.0 mg/L 2,4-
D and after about 1 month into liquid B5 to initiate cell suspensions.

3.5.2 Enhancement of Isoﬂavonoid Synthesis

A possible means to enhance production of isoﬂavonoids in unorganized cell
cultures is through elicitor promotion by phytoalexins (Dougall, 1986), which are
produced in a defensive response to infection by pathogens. Isoﬂavonoids are
intermediates in the biosynthetic pathways of phytoalexins. Elicitation occurs via
activation of genes involved in the biosynthetic pathway (Chappell and Hahlbrock,
1984). This pathway begins with the conversion of phenylalanine to cinnarnic acid by

phenylalanine ammonia-lyase. This enzyme is considered to be a key switchpoint

hm

hm

20

between primary and secondary metabolism of phenolics, and increased enzymic activity
has led to increased product synthesis (Holden and Yeoman, 1987). Additionally, the
supply of phenylalanine has limited capsaicin accumulation (secondary phenolic) in
Capsicum ﬁutescens. The conversion of phenylalanine is apparently important in
regulation of phenolics production, and may thus be an important rate-limiting step in the
pathway.

Several types of phytoalexin elicitors have been studied including fungal cell wall,
plant cell wall, and microbial enzymes (Darvill and Albersheirn, 1984). Hahlbrock et al.
(1981) ﬁrst showed that UV light, a non-biological elicitor, stimulated biosynthesis of
ﬂavonoids. A variety of other abiotic inducers have also been investigated including heavy
metals, autoclaved ribonuclease, chloroform, and detergents, although they are typically
associated with tissue damage (Darvill and Albersheirn, 1984). As isoﬂavonoids such as
genistein and daidzein are intermediates in the production of phytoalexins, production of
these isoﬂavonoids may also be stimulated by phytoalexin elicitors. Kochs and Grisebach
(1987) found that microsomal preparations from normal soybean cells or from elicitor-
challenged cells (P. megasperma or Altemaria carthamz) catalyzed the conversion of (28)-
naringenin to genistein. Miller et al. (1994) showed that the beta-1, 6-1, 3-glucans isolated
from B. japonicum elicited the production of both daidzein and glyceollin in a soybean
cotyledon assay. In a contrasting study, Zacharius and Kalan (1990) studied biotic
stresses on soybean cell cultures with particular reference to quantifying daidzein and
genistein levels (precursors of the phytoalexin glyceollin). Glyceollin levels were generally

promoted while the precursor (daidzein and genistein) concentrations always decreased

21

after 40-68-96 h treatment with the intact bacterial (Pseudomonas syringae pv. glycinea,
strains A-29-2 and 2159) or isolated fungal elicitor (Phytophthora infestans). While
phytoalexin elicitors promoted an enzyme upstream of genistein (Kochs and Grisebach,
1987), the downstream portion of the pathway is apparently also promoted. As a result,
except for daidzein in the Miller et al. (1994) report, there is typically a net decrease in
the concentrations of genistein and daidzein following elicitation. In contrast, cool white
ﬂuorescent light was found to promote genistein accumulation in line Sb-la (Zacharius
and Kalan, 1990). Exposure of Sb-la cells in culture to a synthetic photon ﬂux of 17.5
versus 5.3 umol/mzs, on a 24 h photoperiod, yielded up to ten times more daidzein and
genistein.

Gas-phase reactants are also expected to inﬂuence soybean-cell grth and
product synthesis in the MF B. The importance of carbon dioxide and ethylene in the gas
phase of bioreactors has been documented, but is poorly understood. Addition of 2%
carbon dioxide improved growth and viability of C. roseus in a 20 L airlift bioreactor,
while 5% carbon dioxide was detrimental (Archarnbault, 1991). Both carbon dioxide and
ethylene were important for product formation by Thalictrum rugosum in a 2.3L airlift
bioreactor (Kim et al., 1991). The speciﬁc berberine content was doubled by supplying
these gases at levels comparable to closed shake-ﬂask cultures. Cho et al. (1988) used
ethephon, an ethylene precursor, to stimulate alkaloid production in Coﬂea arabica and
T. rugosum which have widely different biochemical pathways. Gas recycle was found to
increase product synthesis for Catharanthus roseus cultures in a stirred tank reactor

(Schlatrnann et al., 1993). In contrast, carbon dioxide enrichment of sparging air did not

22

enhance growth of Hyoscyamus muticus root cultures (McKelvey et al., 1993). More
recently, Schlatrnann et al. (1997) have suggested a focus on dissolved gaseous
metabolites (DGM), represented by their largest constituent C02, for plant-cell

bioreactor design.

3.6 Diffusion Gradient Chamber
3.6.1 Description

The level of secondary metabolites produced by plant cells is strongly inﬂuenced
by key effector variables such as carbon source, nitrogen source, hormones, precursor
compounds and pH. However, these effects are poorly understood and efforts to increase
production levels typically rely on trial-and-error studies. Because of the large number of
potential effectors (and much larger number of permutations), and slow plant-cell growth
rates, screening studies in shake ﬂasks or petri dishes can be very labor intensive and time
consuming.

A research tool used to-date for microbiological work has great potential to greatly
accelerate these trial-and-error studies with plant cells. The Diffusion Gradient Chamber
(DGC) generates multiple, simultaneous, chemical gradients within a semi-solid growth
medium (Emerson et al., 1994). It consists of a transparent, sterilizable, plastic box that
houses a 5 cm x 5 cm x 1 cm slab-shaped arena containing agar gel illustrated in Figure 5.
The arena is separated from the four liquid reservoirs that bound it by membranes that
allow diffusion of small molecules but are impermeable to gel or cells. A continuous ﬂow
of fresh medium through the reservoirs replenishes solute molecules that diffuse across

the membrane, thus maintaining a constant concentration in the reservoirs. When different

23

  
 
 
 
 
  

 

 

 

R I 1.1 Stud / Reservotr
' ‘ 0". ' rThtnnbscrcw
. ==-~ Outlet
Gas Port
"'=i -~.. Inlet
I

 

 

 

 

 

 

 

 

 

 

 

Figure 5. Diffusion Gradient Chamber.

solute concentrations are used in the reservoirs, diffusion generates a continuous
concentration gradient across the arena. Gradients of multiple solutes may be established
simultaneously, either parallel or perpendicular to one another. In this way, the inﬂuence
of multiple stimuli on cells may be tested simultaneously. Bacterial growth and migration
are measured using a CCD camera and image processing. Applied lighting is diffracted by
cells while a light-absorbing background is placed behind the DGC. Therefore, cells
appear as white on a black background. Bacterial concentrations are directly proportional

to greyscale value over a range of concentrations (M. Widman, unpublished data).

24

3.6.2 Microbial Research

DGC research has centered around the behavior of bacteria in environments
containing chemical gradients. The growth and migratory behavior of E coli. in response to

chemoattractants (aspartate, or-methyl aspartate, and serine) and a chemorepellent

(valine) have been examined by Emerson et al. (1994). Chemoeffector gradients were also
used to separate E. coli from Pseudomonas ﬂuorescens based on differential chemotactic
responses. Bacterial behavior in the DGC has agreed well with predictions of
mathematical models. Fickian diﬁusion in the gel arena and a ﬁrst-order mass-transfer
resistance at the membrane have accurately predicted gradient concentrations (Emerson et
al. , 1994; M. Widman, personal communication). Extensive modeling of bacterial grth
and chemotaxis in response to both consumable and non-consumable substrate gradients
has been completed (Widman et al., 1995). Oxygen and glucose gradients have been
measured utilizing microelectrodes and microbiosensors which allow non-disruptive

assaying (Emerson et al., 1993, Brune et al., 1995).

4. MFB DEVELOPMENT AND CONTINUOUS OPERATION

4.1 Introduction

A wide variety of plant secondary metabolites have value as pharmaceuticals,
food colors, ﬂavors and fragrances. Plant cell culture offers a high-quality, uniform
product while reducing labor costs (Shuler et al., 1990) and eliminating the effects of
weather and disease. Furthermore, continuous production may provide increased
productivity over traditional batch systems.

Bramble et al. (1990) previously demonstrated the use of plant cells in a
magnetically stabilized ﬂuidized bed. Coﬂea arabica cells were co-immobilized with
magnetite in calcium-alginate beads. These beads comprised the solid phase of a two-
phase bed, where solids mixing was prevented by a DC magnetic ﬁeld. Continuous solids
throughput was discussed but not achieved in their system. Solids throughput with
minimal solids mixing is desirable for plant cells as it allows tight control of residence
time. Intracellular products which are difﬁcult or impossible to release into the medium
by permeabilization (Brodelius, 1990) can be extracted from the plant tissue after it has
spent an optimal residence time in the bioreactor.

The goal of this work was to achieve continuous production of daidzein and
genistein from soybean in a magnetoﬂuidized bed (MFB) with continuous biocatalyst
throughput. The isoﬂavonoids are high-value compounds having both medical and

agricultural applications. A magnetic valve for solids (MVS) was used to control the rate

25

26

of solids throughput. Supporting work included batch shake ﬂask experiments to study

the effects of immobilization and antimicrobial agents on soybean culture growth.

4.2 Materials and Methods
4.2.1 Chemicals

All chemicals for soybean culture medium were purchased from Sigma Chemical
(St. Louis, MO) except sucrose from J. T. Baker (Phillipsburg, NJ). LB medium was
purchased from Gibco (Paisley, Scotland). Sodium alginate and genistein were purchased
from Sigma Chemical. Stainless steel powder (grade 410-L) was purchased from Alfa
Aesar (Ward Hill, MA). Daidzein was purchased from ICN Biomedicals (Aurora, OH).
Ampicillin and gentarnicin were purchased from Sigma Chemical, and Timentin was
purchased from SmithKline Beecharn (Philadelphia, PA). Benlate was supplied by
DuPont (Wilmington, DE). Helium was purchased from AGA Gas (Cleveland, OH).
Acetonitrile, ethyl acetate, and acetic acid were purchased from EM Science (Gibbstown,
NJ). All other chemicals were purchased from J. T. Baker.

4.2.2 Equipment
4.2.2.1 MFB System

The MFB system used in this work is illustrated in Figure 6. The main
components of the MFB bioreactor system are the ﬂuidized bed, solenoids, biocatalyst
reservoirs, and mixing vessel. The ﬂuidized bed was comprised of a 60 cm long, 5 cm
diameter Pyrex glass column with grooved ends for O-rings. Custom machined stainless
steel endplates were sealed with Viton O-rings. The head plate, as illustrated in Figure 7,

had a port for solids inlet and a port for liquid outlet. The base plate, shown in Figure 8

27

Gas Feed

 
   
   

Glass Column \

 

 

 

 

 

Guandllmldﬂeoyde

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Fresh Media Product Biocatalyst

Figure 6. Diagram of the magnetoﬂuidized bed bioreactor.

28

had a center outlet for solids and an annular region for liquid inlet. The annular region was
ﬁtted with a stainless steel ﬂuidization mesh (Michigan Dynamics, Heathrow FL) which
provided liquid distribution.

_ Solids
Stainless steel

head plate

\
*‘ﬁ

G
ZZZ 77' 222'

\ Liquid
O-ring
Glass Column /

Figure 7. Cross-sectional view of MFB head plate.

 

 

 

 

 

 

 

 

 

 

A magnetic valve for solids (MV S) was used to control biocatalyst ﬂow ﬁ'om the

MFB, and is conceptually illustrated in Figure 9. An MVS consists of a magnetic coil
Sul'l‘ounding the column in the vicinity of a magnetically susceptible screen. The screen
oDerrings are large enough to allow free passage of solids when no current is applied to the
Valve coil. Supplying power to the valve coil results in an induced magnetic ﬁeld at the
SF-‘rl‘een that stops solids ﬂow. Details of MVS construction and operation are given by
JarBiz-M. et al. (1984) and Yang et al. (1982). The magnetic valve screen, which consisted

of a 430 stainless steel wire mesh 4 x 4 (Belleville Wire Cloth, Cedar Grove NJ), was ﬁxed

29

to the glass with silicone sealant near the bottom of the column. An axial, DC magnetic
ﬁeld was supplied by three solenoids designed to provide an approximately even ﬁeld
strength over the total length of the solenoids. The bottom solenoid also provided the
magnetic ﬁeld for the magnetic valve for solids. Power for the solenoids was supplied by

a Kepco (Flushing NY) DC power supply (ATE 100-5M) rated at 100 volts and 5 A.

Glass
Column Solids

/ Fluidization O-ring
mesh /

 

 

 

 

 

Liquid

 

 

Stainless steel
base plate

Figure 8. Cross-sectional view of MFB base plate.

The feed biocatalyst reservoir was a 2 L aspiration vessel sealed with a foam
stopper and aluminum foil. A Wheaton solvent ﬂask with a modiﬁed screw cap served as
the product biocatalyst reservoir. Biocatalyst outlet tubing passed through the cap via a
hole that was sealed with silicone sealant. Both reservoirs were connected to their
respective headplates with silicone tubing. Clear tubing (1.5 cm ID) allowed visualization

0f biocatalyst movement as it was controlled with pinch clamps. The mixing vessel was a

30

Bioﬂo II (New Brunswick, Edison NJ) fermentation system with a 1.5 L vessel. The
Bioﬂo II system provided agitation, temperature control, and dissolved oxygen
monitoring. Connections to the vessel included a suction line to the baseplate of the
column, a return line from the headplate of the column, a fresh medium inlet, a sparging
gas inlet, and a head space vent. All tubing was norprene or silicone. Liquid was pumped
to the column with a high capacity Masterﬂex peristaltic pump (Cole Parmer, Chicago).
Liquid returned from the column by gravity. A 20 L carboy provided fresh medium via a
peristaltic pump (Cole Parmer) as needed. An oxygen cylinder regulated at 5 psi was
connected to the vessel sparge line. The gas ﬂow rate was controlled by a peristaltic

pump (Cole Parmer).

Glass Column\

 

Solenoid \

    

 

Ferritic . _
Stainless Steel” Solids Liquid .
Screen Distribution

 

 

Figure 9. 3-D diagram of the magnetic valve for solids.

31

4.2.2.2 HPLC

Isoﬂavonoid analysis was conducted using high performance liquid
chromatography (HPLC). A Waters (Milford, MA) 600 multisolvent pump, Waters 490
multiwavelength UV absorbance detector, Waters system interface module, and a PC for
data acquisition and processing comprised the system. A 25 cm by 4.6 mm Rainin

(Wobum MA) Microsorb-MV column containing 5 urn C-18 packing was used. A

reﬁllable 3 cm guard column with pellicular C-l8 packing (Alltech, Deerﬁeld IL) was used
to protect the analytical column.

4.2.2.3 Gas Chromatography

Gas chromatography of the MF B gas phase was performed on a Perkin Elmer
(N orwalk CT) Autosystem Gas Chromatograph with a thermal conductivity detector.
Fixed gases were separated on a 9.1 m by 2.2 mm Hayesep DB 100/ 120 SS column
(Alltech) with helium as the mobile phase. A Perkin Elmer 600 Series Link and PC were
used for data acquisition and instrument control.

4.2.3 Plant Cell Culture and Analysis

Suspension cultures were maintained on a Lab-Line (Melrose Park IL) 3590
Orbital Shaker. All culture transfer and immobilization was conducted in a sterile laminar
ﬂow hood (Baker Company, Sanford ME). A Stir Pak lab stirrer (Cole Parmer) and a
Teﬂon microcentrifuge sample pestle (Fisher, Pittsburgh) were used for extraction of
isoﬂavonoids. Samples were centrifuged in an Eppendorf 5415C centrifuge (Brinkmann
InStruments, Westbury NY). A Cenco drying oven (Central Scientiﬁc, Chicago) was used

for dry weight analysis. All sterilization was performed in a Amsco (Farmington, MI)

32

steam autoclave at 121°C. A Reichert Microstar IV microscope (Cambridge Instruments,

Buffalo NY) was used for light microscopy of biocatalyst. A Pulnix TM—7CN CCD
camera (Sunnyvale, CA) was used for image capture. Image data were acquired using
Quanta (Mountain View CA) WinVisionPro soﬂware on a PC.

4.2.4 Respiration

Dissolved oxygen was measured with an Ingold 0; sensor (12 mm diameter) and
an Ingold Model 170 oxygen ampliﬁer (Mettler-Toledo, Wilmington MA). Data were
acquired with a PC, DT2801 analog and digital I/O board (Data Translation, Marlboro
MA), and Labtech Notebook software (Laboratory Technologies, Wilmington MA). For
suspension culture, the oxygen sensor was inserted into a 100 mL graduated cylinder via a
rubber stopper. For immobilized soybean, a miniature packed bed system was assembled
using a 15 cm by 2.5 cm ID Pyrex column. Column ends were plugged by rubber stoppers
which were connected via Tygon tubing (Cole Parmer). The bottom stopper was ﬁtted
with a metal screen to prevent passage of biocatalyst. The oxygen sensor entered the
column through the top stopper. Liquid was circulated with a Masterﬂex peristaltic
pump (Cole Parmer).

4.2.5 Soybean Culture

Suspension cultures of Glycine max cv. Mandarin were maintained at 25-27 °C on

a rotary shaker at 125 rpm in 250 mL Erlenmeyer ﬂasks. They were subcultured by
passing two-week-old culture through a 400 urn sieve. Twenty mL of this inoculurn was

added to 50 mL of fresh 1-B5 medium as listed in Table 2 (Gamborg et al., 1968). All

"ausfers were conducted in a laminar flow hood. Subculturing had been ongoing for over

33

one year and had produced stable suspension cultures. While undocumented, a decrease in
average aggregate size (ﬁner suspensions) and increase in average growth rate have been
observed. Both results are advantageous for productivity and immobilization.

Table 2. Composition of soybean medium.

 

 

It_em Concentration
Gamborg B5 Salts 3.1 g/L
Nicotinic acid 1 mg/L
Pyridoxine HCl 1 mg/L
Thiamine HCl 10 mg/L
Myo-inositol 100 mg/L
Sucrose 20 g/L

2,4-D 1 mg/L
Kinetin 0.15 mg/L

 

 

 

4.2.6 Immobilization

Immobilization solution (150 mL aliquots) consisting of 2% medium viscosity
sodium alginate and 25% (w/w) stainless steel was prepared and autoclaved in 500 mL
Erlenmeyer ﬂasks. A soybean inoculum was prepared by passing an approximately 2
week-old (since last sub-culture) culture through a 1 mm sieve. The ﬁltrate was added to
the immobilization solution and peristaltically pumped through a sterile bead-making
apparatus into hardening solution. Figure 10 illustrates the section of the beadmaker
where formation of beads occurred. The bead-making apparatus consisted of feed tubing
leading to four 16 gauge syringe needles with ﬂattened tips. Each needle was ﬁtted with

concentric tubing to provide a shearing stream of sterile ﬁltered air that decreased bead

34

size. Beads were approximately 2 mm in diameter. Hardening solution (400 mL) consisted
of 0.1 M calcium chloride and 20 g/L sucrose. After 20 minutes, the hardening solution
was decanted and replaced with 200 mL of l-BS medium. For MFB use,
antibiotics/antimycotics were added (50 mg/L ampicillin, 100 mg/L Timentin, and 25 mg/L
Benlate). Before addition to the MF B, each batch of fresh biocatalyst was sampled
(approximately 10 mL) with a sterile 25 mL wide-mouth pipette for dry weight analysis.
The entire process was conducted in a laminar ﬂow hood.

lm mobilization

Matrix

1 Sterile Air

Norprene tubing

Spatial arrangement of
needles as viewed from
bottom.

©©
©©

 

  

 

\ 16 G needle

Figure 10. Beadmaker for co-immobilization of soybean and stainless steel.

35

4.2.7 Cell Recovery

Biocatalyst volume was measured by volume displacement in a 50 mL graduated
cylinder. Sample volumes were approximately 10 and 20 mL for feed and product
samples respectively. Samples were placed in 125 mL Erlenmeyer ﬂasks with
approximately 50 mL of 0.2 M sodium citrate to dissolve the alginate matrix. Flasks were
placed on a rotary shaker for 30 minutes at 300 rpm. Suspended cell mass was decanted
while stainless steel was retained by placing a magnet on the bottom of the ﬂask.
Complete recovery of cell mass was achieved by removing the magnet, adding deionized
water to wash the stainless steel, and decanting suspended cell mass with the magnet on
the bottom of the ﬂask.

4.2.8 Analytical Procedures
4.2.8.1 Dry Weight

Decanted cell mass was collected by ﬁltering through dried and tared Whatrnan 41
ashless ﬁlter paper using a glass microanalysis ﬁlter holder (Fisher, Pittsburgh PA).

Samples were dried ovemight at 70°C and weighed. Samples immobilized with stainless

steel were then ashed in tared, ceramic crucibles over a Bunsen burner, cooled in a
desiccator, and weighed. The ash in the crucible was taken as the mass of residual
stainless steel which was not separated from the soybean tissue.

4.2.8.2 Biocatalyst Microscopy

Product biocatalyst was sliced to obtain a cross-section ﬁ'om the center of the

Particle. The cross-sections were examined at 40 times magniﬁcation for distribution of

36

cell mass. Both transmission and dark-ﬁeld illumination were utilized. Images were
acquired by mounting a CCD camera on the microscope.

4.2.8.3 Extraction of Isoﬂavonoids

Soybean tissue recovered from product biocatalyst was extracted by the following
method adapted from Graham (1991). Product biocatalyst was stored frozen until used.
Approximately 0.200 g of soybean tissue was extracted twice with 0.4 mL of 80%
ethanol (20% water) in a 1.5 mL rnicrofuge tube by grinding with a Teﬂon microfuge
pestle for 5 minutes. Previous testing veriﬁed a greater than 95% average recovery of the
isoﬂavonoids using this extraction protocol. The combined extracts were injected directly
into the HPLC after centrifugation. Following HPLC analysis of the extracts, a portion of
each sample was subjected to a hydrolysis treatment to liberate daidzein and genistein
from sugar moiety compounds. In a screw-top test tube, 0.4 mL of sample was combined

with 4 mL of 1 N HCl and put in a 100 °C water bath for 1 hour. The hydrolysis mixture
was then extracted with 1 mL of ethyl acetate. The ethyl acetate extract was dried at 40°C

in a rnicrofuge tube and dissolved in 0.4 mL methanol (ethyl acetate adversely affected
chromatograrns). The hydrolyzed extracts were analyzed by HPLC in the same manner as
raw extracts.

4.2.8.4 Isoflavonoid Analysis

HPLC analysis was performed with gradient prograrnrrring. Mobile phase A was
acetonitrile with 0.5% acetic acid and mobile phase B was water with 0.5% acetic acid.
Adjustment of pH with acetic acid signiﬁcantly reduced band tailing. The gradient linearly

increased ﬁ'om 30% A to 45% A over 20 minutes, followed by a 2 minute linear increase

37

to 90%A, a 10 minute hold, and a 5 minute linear return to 30% A. Retention times were
8.0 and 14.9 minutes for daidzein and genistein respectively. UV absorbance was
monitored at 254 nm.

Identities of daidzein and genistein where veriﬁed by sample spiking and mass
spectrometry. Retention times of daidzein and genistein in the sample matrix were
checked by adding standards to soybean extracts. Fractions of the HPLC efﬂuent were
collected and analyzed by electrospray mass spectrometry in negative ion mode (M- 1 ).
Daidzein and genistein were veriﬁed by peaks at 253 and 269 amu, respectively.

4.2.8.5 Respiration

For suspension culture, the entire contents of a 250 mL shake-ﬂask culture were
put into the graduated cylinder along with fresh medium to approximately 100 mL total
volume. The liquid was then sparged with air for 5 minutes. The oxygen sensor was
inserted with care to remove any head space, and the stirring speed was set to give
complete mixing in the cylinder. Dissolved oxygen concentration data were collected until
a steady value was reached.

For immobilized culture, the apparatus described above was ﬁlled with product
biocatalyst. The oxygen sensor was inserted, and the remaining volume was ﬁlled with
deionized water. The pump speed was set to keep the liquid well mixed.

4.2.9 Batch Shake Flask Cultures

Shake ﬂask cultures were conducted to study the effects of immobilization on
soybean growth. The study compared suspension culture, cells immobilized in alginate,

and cells immobilized in alginate with 25% (w/w) stainless steel. Replicate cultures in l L

38

Erlenmeyer ﬂasks contained 400 mL of l-B5 medium. Suspension cultures had 30 mL of
inoculum while immobilized cultures had 20 mL of inoculum in approximately 200 mL of
beads. A common inoculum was used for all ﬂasks to reduce error due to soybean
variation. Immobilization followed the procedure described above. Cultures were
maintained on an orbital shaker at 125 rpm.

Shake ﬂask cultures were also conducted to study the effects of antimicrobial
agents on soybean growth. Effects of ampicillin (100 mg/L), gentarnicin (50 mg/L),
Timentin (200 mg/L), and Benlate (50 mg/L) were studied in 1 L Erlenmeyer ﬂasks
containing 400 mL of 1-B5 medium. A common inoculum (30 mL) was used for all ﬂasks,
including two controls.

4.2.10 MFB Operation
4.2.10.1 Start-up

As the entire MFB system was designed to ﬁt in the autoclave, system
connections made after autoclaving were minimized. Only the two connections to the
column headplate were made post-autoclaving as necessitated by the solenoids. The fresh
medium reservoir was also ﬁlled prior to autoclaving and did not contain
antibiotics/antimycotics. Aﬁer mounting and ﬁnal connections were made, the system
was ﬁlled ﬁ'om the medium reservoir and liquid circulation started. The liquid ﬂow rate
was set to approximately 1.6 L/min. Ampicillin, Timentin, and Benlate were added to the
circulating medium yielding 50, 100, and 25 mg/L concentrations respectively. The
solenoid power supply was set at 2 A, yielding a magnetic ﬁeld strength of 150-180 gauss

depending on position in the column. The column was loaded with immobilized soybean

39

(feed biocatalyst) in four aliquots over the ﬁrst three days of operation. Agitation in the
Bioﬂo vessel was set at 250 rpm. Sparging was not started until the dissolved oxygen
concentration dropped below 80% of the initial value.

4.2.10.2 Continuous Biocatalyst Processing

At intervals varying from one to several days, product biocatalyst was removed
from the bottom of the column, and feed biocatalyst was added to the top. Intervals
during the ﬁrst 12 days were chosen with the goal of demonstrating growth during that
period. From day 12 through 34, scheduling obligations resulted in irregular biocatalyst
addition/removal. Past day 34, biocatalyst was exchanged roughly every four days. To
minimize bed mixing, the liquid ﬂow rate was reduced to 1 L/min before biocatalyst
removal operations. The power supply to the solenoids was increased to 3 A. The upper
of two pinch clamps on the outlet tubing was opened slightly to allow the length of
tubing between the upper and lower pinch clamps to ﬁll with medium from the column.
At the same time, air was displaced into the column. Once the tubing was ﬁlled with
liquid, the upper clamp was completely removed. The magnetic valve was operated by
decreasing the solenoid power to 0.6 A which reduced the magnetic interaction between
the biocatalyst and screen. The magnetic ﬁeld strength was approximately 40 Gauss at
the magnetic valve. Squeezing the outlet tubing induced the biocatalyst to fall through the
stainless steel screen into the region of the column above the ﬂuidization mesh. Once
entering this region of free movement, the biocatalyst fell through the center opening in
the baseplate. No mixing in the bed was observed due to the stabilization afforded by the

50-60 gauss magnetic ﬁeld in the column.

40

After the desired quantity of beads had passed through the screen, the solenoid
power was returned to 3 A. While the greater portion of the released biocatalyst fell
through the outlet hole in the baseplate, the remainder which came to rest on the
ﬂuidization mesh was removed by brieﬂy increasing the liquid ﬂow rate to 3 L/min. Once
all the biocatalyst was in the outlet tubing, the liquid ﬂow rate was returned to 1.6 L/min.
The bottom clamp was then removed, releasing the biocatalyst into the product reservoir.
The bottom clamp was replaced, and the process repeated once more, resulting in 150-
200 mL of biocatalyst in the reservoir. Following biocatalyst removal, additional medium
was drained into the product reservoir making room for addition of fresh medium. The
solenoid power was returned to 2.0 A. The ﬁnal step in removal was a aseptic exchange
of the full biocatalyst reservoir with an empty one.

After product biocatalyst was removed, feed biocatalyst was added using a similar
pinch-clamp technique. Freshly prepared biocatalyst and medium were added to the feed
biocatalyst reservoir. The upper of two pinch clamps was removed allowing fresh
medium and biocatalyst to ﬁll the inlet tubing. Repeated squeezing of the tubing
facilitated movement of solids ﬁom the reservoir to the tubing, displacing the liquid. Once
the tubing was full of biocatalyst, the top pinch clamp was replaced and the bottom pinch
clamp removed allowing the biocatalyst to enter the column through the headplate. The
bottom clamp was replaced and the process repeated until all biocatalyst was removed
from the reservoir (4-5 times). Fresh medium was then added (approximately 200 mL) by
releasing the pinch clamps. The pinch clamps were then replaced, completing the

removal/addition process.

41

Aseptic conditions during continuous operation were veriﬁed by periodically
inoculating a rich medium (LB broth) with MFB medium from the product reservoir.
Both LB agar in petri dishes and LB medium in 125 mL Erlenmeyer shake ﬂask cultures
were employed. Shake ﬂasks were incubated at room temperature while petri dishes were
incubated at both room temperature and 30°C for a minimum of two weeks.

4.2.11 Theoretical

4.2.11.1 Speciﬁc Growth Rates

Growth rates during continuous operation were estimated ﬁ'om residence times

and cell densities in the feed and product biocatalyst. For exponential growth, the average

speciﬁc growth rate ([1) is described by
( X ] ( 1 )
1n —
X0

where 1' is the biocatalyst residence time, and X and X0 are the soybean concentrations in

 

ﬂ:

“1

the product and feed biocatalyst, respectively. Residence times were calculated from the
known volumes and times of biocatalyst addition and removal, assuming plug ﬂow of the
solid phase. Uncertainty in residence times stems from error in measurement of
biocatalyst volume added and removed as well as any mixing which may have occurred in
the column. As error in residence time calculations may lead to erroneous pairing of X and
X0 values, feed biocatalyst cell densities (X0) were taken as the averaged value over the

entire run.

42

4.2.11.2 Thiele Modulus Calculations

The possibility of oxygen diffusion within the biocatalyst limiting the overall

reaction rate was assessed with Thiele modulus calculations. An observable modulus (<13)

has been described by Bailey and Ollis (1986) for Michaelis-Menten kinetics which does

not require knowledge of intrinsic-rate parameters:

2
ch = ”0 E (2)
D350 AP

where v0 is the observed volumetric reaction rate, D, is the diffusion coeﬁicient of

 

substrate in the biocatalyst, s0 is the bulk substrate concentration, Vp is the biocatalyst
volume, and Ap is the biocatalyst surface area. For observable modulus calculations,
irnmobilized-soybean respiration rates were measured as described above. Volume and
area parameters assumed a 2 mm diameter sphere. An oxygen diffusion coefficient of 1.91
x 10.5 cmz/sec was calculated by multiplying the diffusion coefficient in water at 25°C
(Perry and Green, 1984) by correction factors of 0.85 for 2% calcium alginate gel (Westrin

and Axelsson, 1991) and 0.9 for 25% w/w magnetite (Thompson, 1993). The second

correction factor assumes the effects of magnetite and stainless steel at 25% w/w on the
effective diffusion coefﬁcient to be equal. The bulk oxygen concentration at 25°C was
taken to be 8.43 mg/L (Bailey and Ollis, 1986).

The possibility of oxygen diffusion limitations within individual soybean

aggregates was assessed with the generalized modulus:

 

¢_YL n+1kaf’l (3)
A, 2 D,

where k, is the intrinsic rate constant, C, is the bulk substrate concentration, D, is the
effective diffusivity, and n is the reaction order (F roment and Bischoff, 1979). The
intrinsic respiration rate was assumed to be that of suspension cultures, and measured as
described above. The effective diﬁusivity was taken as 2% of the value for oxygen in
water (Ananta et al. , 1995). Respiration rates measured on a mass basis were translated to
a volumetric basis assuming a cell density of 1.02 g/mL (Ananta et al., 1995) and a dry

weight to ﬂesh weight ratio of 0.07 g/g (unpublished data).

4.3 Results and Discussion
4.3.1 Batch Shake Flask Cultures
4.3.1.1 Immobilization

Speciﬁc growth rates for suspension (control), alginate immobilized, and
alginate/stainless steel immobilized cultures were 0.35i0.01, 0211008, and 0.26:l:0.06

day'l respectively. All cultures exhibited exponential growth with r2 values ranging from
0.977 to 0.999 for ﬁtted exponential curves. Figure 11 shows the results for replicate,
alginate immobilized cultures. Batch shake ﬂask cultures indicated immobilization slows
soybean growth. Immobilization with stainless steel reduced the growth rate by about
25% compared to suspension culture. However, the addition of stainless steel in the
immobilization matrix did not further hinder growth compared to alginate only. While the
average speciﬁc growth rate for stainless steel cultures was higher than for alginate only,

the difference was not statistically signiﬁcant at the 95% conﬁdence level.

44

1 .00E-02

 

Concentration (glmL)

 

 

1.00E-03 : i i i i
6 8 10 12 14 16 18

 

Time (days)

Figure 11. Immobilized soybean cell densities for replicate shake ﬂasks.

Three aspects of the immobilization protocol merit discussion. First, this work
used fenitic stainless steel, a novel material for creating a magnetically susceptible
biocatalyst, in place of magnetite. The stainless steel powder chosen for this work offers
several advantages over magnetite including a higher magnetic permeability, inherent
biological inertness, and easier separation from dissolved immobilization matrix due to
increased particle size and density. Stainless steel concentrations employed in this study
were much higher than magnetite concentrations reported previously (Bramble et al. ,
1990), and considered realistic for stable bed operation with continuous solids
throughput.

Second, the use of concentric air streams to shear immobilization matrix droplets
from needles was valuable in maintaining a small biocatalyst diameter. The use of
concentric, shearing air reduced bead diameters by roughly 50%. Smaller diameters reduce

diffusional mass transfer limitations. The minimum size is limited, however, by the

45

relatively large size of suspension culture aggregates. In order to utilize the majority of a 2
week-old culture for inoculum, a 1 mm screen was used. This mesh size sets the minimum
practical needle size at about 16 gauge.

Third, experimental evidence suggests that stainless steel became entrapped within
soybean aggregates as they grew. While separating soybean cells released from the alginate
matrix ﬁ'om stainless steel, some soybean aggregates appeared to be drawn to the magnet.
Tissue resting on the side of a ﬂask could be moved with a magnet. It was unclear
however, if the movement was due to entrapped metal or the presence of metal around
the aggregates. Examination at 100 times magniﬁcation showed steel particles on or in the
aggregates (results not shown). Again, a clear distinction could not be made. Finally,
complete separation of soybean cells and stainless steel was not achieved following
dissolution of the alginate, as determined by ashing of dry weight samples. Due to the
density of stainless steel, it was necessary to account for the mass of steel when
determining cell dry weights.

4.3.1.2 Antimicrobial Agents

Table 3 lists the antibiotic-antimycotic compounds tested, concentrations used
and resulting speciﬁc growth rates. Of the four agents tested, three were found to have
limited toxicity at concentrations useful in inhibiting microbial growth. Only gentarnicin
was found to be completely toxic to plant cells at the tested concentration. The remaining
compounds offer a broad spectrum cocktail to prevent contamination during extended,
continuous bioreactor operation. Ampicillin is a broad range antibiotic effective mainly

against Gram positive bacteria (Pollock et al., 1983). Timentin contains ticarcillin,

46

effective against Gram positive and negative bacteria, as well as clavulanic acid, an anti-
betalactarnase. Timentin therefore bolsters the stability of ampicillin. Benlate is a broad
spectrum antifungal agent (Brown et al., 1982).

Table 3. Antibiotics successfully tested with soybean cell culture.

 

 

58931 M Speciﬁc Growth Rate (day'l)
Control - 0.40:l:0.13

Ampicillin 100 0.37

Benlate 50 0.32

Timentin 200 0.3 l

Gentamicin 50 no growth

 

 

 

4.3.2 Asepsis Testing

Trial runs of the MFB were conducted to verify aseptic operation. Initially, the
column and endplates assembly was partially ﬁlled with LB medium, sealed, and
autoclaved. No microbial growth was apparent after 8 days at room temperature. The
complete MFB system was then tested including liquid circulation and air sparging using
LB medium. Asepsis was veriﬁed by plating of medium on LB agar after 4 and 6 days of
operation. A ﬁnal trial nm was made to verify that biocatalyst could be added and
removed from the system aseptically. Alginate beads containing 25% stainless steel were
added/removed four times over 24 days without contamination as veriﬁed by LB plating

of the medium. No antimicrobial agents were used in asepsis testing.

47

4.3.3 Continuous Operation

4.3.3.1 Soybean Growth

Continuous operation of the MF B utilizing immobilized soybean culture was
accomplished during a 62 day run. No evidence of contamination was seen on LB petri
dishes or shake ﬂasks. The product biocatalyst loadings are shown as a function of time
in Figure 12. Over the ﬁrst two weeks, the goal of operation was demonstration of
soybean growth in the MFB system. Residence times were therefore kept short, and
process monitoring was limited to product biocatalyst cell densities and dissolved oxygen
concentrations. Figure 12 shows that cell densities increased over the ﬁrst 12 days, in part
due to increasing residence times during start-up. The limited oxygen demand was met by
sparging the liquid each day until the dissolved oxygen level returned to the air saturation

level.

12

 

 

 

 

U

10~-
E: 3-. o
2% s-
C
8
= 4“
8

O
2 ° 0
0 Ga 1_ 1 ﬁr to 043

o 10 20 3o 40 50 60 70
Time (days)

Figure 12. Product biocatalyst loading as dry weight per biocatalyst volume.

48

Following start-up was a period of irregular biocatalyst addition/removal. During
this time, residence times were increased, and additional monitoring methods were
developed. Daily (periodic) sparging was replaced with continuous, peristaltic-pump
addition of oxygen to the Bioﬂo vessel to match the increased oxygen uptake. As seen in
Figure 12, only two product biocatalyst aliquots were removed between days 12 and 34.
Soybean concentration in the product biocatalyst peaked at day 34. The signiﬁcant
decrease in growth after day 34 is likely explained by two experimental problems that are
described below.

On day 30, agitation in the Bioﬂo vessel ceased due to bearing failure. Gas-liquid
mass transfer was therefore hindered for the remainder of the experiment. Within three
days after agitation ceased, dissolved oxygen levels dropped considerably below air
saturation levels. An effort to counter decreasing oxygen levels was made by gradually
increasing oxygen feed to the Bioﬂo vessel. However, dissolved oxygen levels continued
to drop through day 45 while head space analysis indicated oxygen levels well above
ambient conditions (results not shown). During the same time, product biocatalyst
loadings in the MFB decreased signiﬁcantly. These results support the hypothesis that
loss of agitation resulted in poor gas-liquid mass transfer, depletion of oxygen in the
medium, and poor soybean growth.

A second cause of poor growth after day 34 may be dilution of endogenous
growth factors (Bramble et al., 1990). During biocatalyst removal on day 34, the soybean
culture was found to be growing out of the biocatalyst at the bottom of the MFB. Some

biocatalyst particles were connected by plant tissue, making the removal procedure

49

difficult. Growth outside of the immobilization matrix also resulted in high medium
turbidity. Between days 34 and 38, 2.4 L out of the approximately 3 L working liquid
volume was exchanged with ﬁesh medium to counter this problem. While soybean growth
in the liquid phase was avoided, dilution of important plant hormones may have reduced
growth rates.

Figure 13 shows speciﬁc growth rates as a function of time. Growth rates mirror
product biocatalyst densities as expected. With caution regarding uncertainties in
residence time calculations, the MFB growth rates can be compared to batch shake ﬂask
cultures of immobilized soybean. The peak MF B value of 0.17 day'1 compares well to
the shake ﬂask value of 0.26 day'l. The average MFB growth rate is roughly one-third of
that in shake ﬂask culture. Furthermore, at least part of the lower growth rates can be
explained by stress due to the antibiotic cocktail used in the MFB, as illustrated by batch
shake ﬂask studies. Growth rates fell to their lowest values after day 50, but showed a
rebound in the biocatalyst removed on the last day of operation.

4.3.3.2 Respiration and Oxygen Diffusion

Thiele modulus calculations suggested that soybean respiration was not rate
limited by oxygen diﬁusion in the biocatalyst. In all respiration measurements, the
oxygen-uptake rates were found to be independent of dissolved oxygen concentration.
The respiration rate of soybean in MF B product biocatalyst was measured at 8.9 mg 02/g

cells-hr. The observed volumetric rate was 2.9 x 10'9 g Oz/mL biocatalyst-sec yielding an

observable modulus of 0.02. A graphical solution of n as a function of (I) is available for

Michaelis-Menten kinetics in the zero-order regime (Bailey and Ollis, 1986). The

50

resulting effectiveness factor of 1.0 indicates that oxygen diffusion within the biocatalyst
is not rate limiting. The immobilized cell respiration rate was roughly one-third of that for
suspension culture in exponential growth phase, 28 mg Ozlg cells-hr. The difference is
consistent with the growth rate of the immobilized culture being approximately 0.1 day'1
(roughly one-third of suspension culture). Furthermore, the immobilized soybean
residence time was approximately 16 days. The respiration rate of a post-exponential
phase suspension culture (19 days old) was 8.7 mg 02/g cells-hr, which is very close to

that of the immobilized soybean culture.

 

 

 

 

0.2
0
€‘ 0.15 4 ° 0
‘9. o
m
B
O
to 0.1~
O!
c o
E, 0.05 —~ °
0 O
o
0 t a i 1 :0 04.
0 10 20 3o 40 50 60 70
Time (days)

Figure 13. Soybean speciﬁc growth rates in the MFB.

Microscopy of product biocatalyst cross-sections supported the idea that oxygen
diffusion was not rate limiting. Typical pictures of biocatalyst cross-sections are shown
in Figure 14. Figure 14a shows the dark-ﬁeld image of a head with soybean aggregates

appearing as lighter regions. Figure 14b shows the same bead under transmission

51

illumination with soybean appearing as darker regions in the center, and stainless steel
preventing passage of light around edges. Figure 14c shows a different bead with three
dark regions of soybean, illustrating larger aggregate sizes. No radial dependence of cell
concentration was observed in the biocatalyst. Soybean aggregates appeared to be
randomly distributed throughout the biocatalyst, as opposed to preferentially located
near the biocatalyst surface. The latter trend was observed by Bramble et al. (1990) with
C. arabica in 3-4 mm alginate beads. Images from days 38, 42, and 62 showed soybean
aggregates to be large relative to suspension culture. Aggregates up to 1 mm across were
seen, with aggregates ﬁ'om 0.1 to 0.5 mm being most common. The general observation of
large and dispersed aggregates may be explained by immobilization technique and
hormone dependent growth. Large aggregates are expected to develop from inoculum
aggregates which are above some minimum size. Growth of such aggregates may be
explained by the feeder and dividing aggregate model proposed by Shuler (1981). Thus,
aggregates which started large may continue increasing in size while smaller ones grow
poorly or not at all. Poorly growing aggregates may however, be bolstered if they are
surrounded or engulfed by rapidly growing large aggregates. Furthermore, the somewhat
dispersed nature of the soybean in the biocatalyst suggests that higher inoculation
densities would provide higher ﬁnal densities for a given residence time. Higher
inoculation densities could reduce the void volume observed in the cross-section images.
While intrabiocatalyst oxygen diffusion does not appear to be rate limiting, oxygen
diffusion within soybean aggregates may be important. Modulus values for typically

observed aggregates (0.1 to 0.5 mm diameter) range from 0.5 to 2.3. Based on a modulus

52

of 1.0, rate limitation due to oxygen diffusion would be expected for aggregates larger than
0.2 mm. For zero order kinetics where o is greater than unity, T] is equal to (1)"1 (Froment

and Bischoff, 1979). The effectiveness factor for 0.5 mm aggregates is therefore 0.4. For
the largest aggregates observed (roughly 1 mm), the calculated effectiveness factor is 0.2.
These results indicate that soybean aggregates may become diffusion limited as they move
through the MFB. However, Ananta et al. (1995) found that experimental results do not
agree with Thiele modulus calculations. Aggregates ranging from 3.0 to 12.5 mm in
diameter were capable of active growth and respiration with retention of cell viability in at
least 90% of the particle; observed speciﬁc oxygen uptake rates were within the range

normally associated with plant cell suspensions. It is therefore unclear whether oxygen

diffusion limits growth rates inside soybean aggregates.

 

Figure 14. Product biocatalyst cross-sections at 40 times magniﬁcation.
4.3.3.3 Isoﬂavonoid Production
Daidzein and genistein were produced during the entire run with concentrations
ranging from approximately 10 to 200 pg per gram of dry soybean tissue. The daidzein
and genistein concentrations are shown as a function of time in Figure 15. The data are for

hydrolyzed extracts, representing the total genistein or daidzein found in the soybean as

53

both free isoﬂavonoid and isoﬂavonoid with moieties. Acid hydrolyzed extracts averaged
nearly 2.9 and 1.5 times the amount of daidzein and genistein respectively in raw extracts.
Thus nearly two-thirds of the daidzein and one-third of the genistein exists with sugar
moieties. The extraction protocol was also tested using two week-old soybean suspension
cultures, providing isoﬂavonoid concentrations for comparison. Concentrations were 340

i 50 ppm and 560 :l: 160 ppm at a 95% conﬁdence level for daidzein and genistein

respectively. Daidzein concentrations in the MFB reached over one-half the average
suspension culture level. Genistein concentrations reached nearly one-ﬁfth the average
suspension culture level. While peak isoﬂavonoid levels in the MFB were signiﬁcantly
below suspension culture values, the results are for the ﬁrst continuous run attempted.
Improved operating procedures and process control can reasonably be expected to raise

the product concentrations considerably.

 

 

 

 

 

 

 

200 5
oDaidzeinj °
. . l
E 150 H DGenisteIn !
a.
e
O O D
g 100 ~~ u
9 D a
o
g 50 «i- % O n
o D a 1::
O D a B
0 i .L t I. . i
0 1O 20 30 40 50 60 70
Time (d)

Figure 15. Isoﬂavonoid content of product soybean as mass per dry tissue.

54

4.3.4 System Improvements

The two month, continuous run demonstrated the feasibility of continuous
production of plant secondary metabolites in an MFB. It also identiﬁed potential
improvements in the experimental protocols that could yield steady-state operation. The
ability to maintain steady-state growth and product accumulation is a likely prerequisite
for further development toward a commercial process.

Automatic control of the dissolved oxygen concentration would help provide
steady conditions for the plant cells. A gas recycle 100p would be able to maintain good
mass transfer between the gas and liquid while minimizing stripping of volatiles that may
regulate the cellular metabolism. A controller could be used to adjust the total gas ﬂow
rate (fresh air) to match the oxygen demand. A high recycle gas ﬂow rate and impeller
speed should be able to maintain a high gas-liquid mass transfer coefﬁcient independent of
the total gas ﬂow rate. This scheme would avoid volatiles stripping by excessive total gas
ﬂow rates typically used to maintain reasonable gas-liquid mass transfer.

In previous operation, liquid medium was added/removed along with biocatalyst
addition/removal. For tighter control of net medium ﬂow, fresh medium could be
continually added to the mixing vessel, and removed via gravity through an overﬂow port.
Improved solids metering during addition and removal could be achieved by use of
graduated tubing for biocatalyst volume measurements. Evacuation of air from the
biocatalyst outlet tubing should avoid large air bubbles rising through and disrupting the
bed. Improved metering and decreased bed mixing would in turn lead to tighter residence

time control.

55

4.4 Conclusions

Continuous isoﬂavonoid production from soybean with biocatalyst throughput
has been demonstrated in a MF B bioreactor for over 2 months. Immobilization of
soybean cell culture in alginate with stainless steel reduced the speciﬁc growth rate about
25% relative to suspension culture. Soybean growth rates and product concentrations

were lower in the MFB than batch suspension cultures. Genistein and daidzein

production ranged from 10 to 200 pg per gram of dry soybean tissue. Hydrolysis of

bound daidzein and genistein was found to be important for product recovery. Both
Thiele modulus calculations and biocatalyst microscopy suggested that intra-biocatalyst
oxygen diffusion was not rate limiting. Ampicillin, Timentin, and Benlate had limited
phytotoxicity at concentrations useful in preventing contamination during long,

continuous bioreactor operation.

5. PROTOCOL AND MODELING FOR LONG-TERM MFB
OPERATION

5.1 Introduction

Several reactor schemes are available for plant cell cultures, including stirred tanks,
bubble columns, air-lift reactors, hollow-ﬁber membranes, liquid-dispersed trickle and
incline reactors (van Gulik et al., 1994; Reinhard, 1989; Panda et al., 1989; Andrews,
1988). Each of these reactors is subject to trade-offs in advantages such as simplicity or
effective mass transfer, and limitations such as shearing of cells, insufficient mixing at high
cell densities, settling of cells, and difﬁculty in scale-up (Panda et al., 1989). Bioreactor
design also requires consideration of characteristics unique to plant cell culture including
the importance of medium conditioning (Bramble et al., 1990), dissolved gaseous
metabolites (Schlatrnann et al., 1997), and difficulty in intracellular product recovery
(Brodelius, 1990).

The magnetoﬂuidized bed (MF B) bioreactor addresses several of the important
issues for a plant-cell bioprocess. This bioreactor scheme was ﬁrst introduced in a batch
mode by Bramble et al. (1990). The authors co-immobilized Coﬂea arabica cells with
magnetite in calcium-alginate beads. These beads comprised the solid phase of a two-
phase bed, where solids mixing was prevented by a DC magnetic ﬁeld. Continuous solids
throughput in a similar system was achieved by Ames and Worden (1997) in a 2 month
run. The residence time of immobilized soybean cell culture was tightly controlled
independent of medium residence time. The need for permeabilization was avoided by

extracting soybean tissue for isoﬂavonoids after harvesting from the reactor.

56

57

While the MFB has been demonstrated for continuous production of intracellular
metabolites ﬁ'om plant cell culture, a framework for establishing operational protocols has
not yet been developed. This work couples plant cell metabolism and bioreactor
characteristics of the MF B system. The MP B has been operated using soybean cell
culture, with the goal of understanding steady-state operation. The system has been
described with appropriate mass balances, an unstructured metabolic model, and

metabolic parameters drawn primarily ﬁ'om the literature.

5.2 Materials and Methods
5.2.1 Chemicals

Chemicals used for culture medium, immobilization, and analytical work have been
reported in Section 4.

5.2.2 MFB System

The MFB system used in this work is illustrated in Figure 16. Details of the
material and construction for the ﬂuidized bed, reservoirs, and mixing vessel (Bioﬂo II,
New Brunswick, Edison, NJ) have been reported in Section 4. Modiﬁcations to the
original system included a gas phase recycle loop, dissolved oxygen control, a vacuum
port on the biocatalyst outlet tubing, and drain lines from the feed biocatalyst reservoir
and mixing vessel (see Figure 16). The gas phase recycle loop provided medium sparging
independent of the feed gas ﬂow rate. Gas was recirculated ﬁom the head space through
the liquid return line to a ﬁitted glass sparger via a peristaltic pump (Cole Parmer,
Chicago). Head space samples were taken from this recirculation loop. The dissolved

oxygen control loop consisted of an Ingold 0; sensor (12 mm dia.) with Model 170

58

oxygen ampliﬁer (Mettler-Toledo, Wilmington, MA), an IBM XT with a DT2801 analog
and digital I/O board (Data Translation, Marlboro, MA) running Labtech Notebook
(Laboratory Technologies Corporation, Wilmington, MA), and a peristaltic pump (Cole
Parmer). The pump was ﬁtted with two heads for 7.9 and 1.6 mm inside diameter tubing
for air and oxygen, respectively. The outlet pressures for the air and oxygen cylinders
were regulated at 35 kPa. The port added for vacuum provided removal of air in the
biocatalyst outlet tubing which would otherwise rise through the magnetoﬂuidized bed,
potentially disrupting the particles. The drain from the feed biocatalyst reservoir was
used to remove medium which was not used in transfer of biocatalyst into the column.
The drain line from the mixing vessel maintained constant medium volume by overﬂow.
The system required only two connections at the column headplate to be made post-
autoclaving.

5.2.3 Soybean Culture and Medium

Soybean suspension cultures were Glycine max cv. Mandarin (Section 4). The
liquid medium was Gamborg BS supplemented with 1.0 mg/L 2,4-dichlorophenoxyacetic
acid (2,4-D), 0.15 mg/L kinetin, 50 mg/L ampicillin, 100 mg/L Timentin, and 25 mg/L
Benlate.

5.2.4 Batch Shake Flask Cultures

Shake ﬂask cultures were conducted to study the effects of the
antibiotic/antimycotic cocktail on isoﬂavonoid production. The study compared the
medium described above to a control medium which contained no ampicillin, Timentin or

Benlate. Replicate cultures in l L Erlenmeyer ﬂasks contained 400 mL of medium and 30

Sample 1 or GC<—I>4—
[ BIOFLO Ii CONSOLE ----------------

Temp. & Agitation ]- . -® -
PC w/ D/A l0 CARD -- _.
D0 Control

 

 

 

 

Feed Gas
Pumps

 

AIR OXYGEN

 

 

 

 

 

 

 

 

FEED
MEDIUM
RESERVOIR

Cooling
(Biofio control)

Sterile
Filter

59

Gas Recirc. Pump

 

 

t—e

FEED
BIOCATALYST

N—><
Gas $-

I.
I

 

 

 

Liquid

 

 

 

  
  

 

 

WASTE
HEDIUM
RESERVOIR

 

 

 

 

(Gas )

   

Glass / -
Column

Magnetic—
Valve

 

 

~©

Medium
Recirc.
Pump > <

Vacuum

 

PRODUCT
BIOCATALYST

 

 

Figure 16. Diagram of the magnetoﬂuidized bed bioreactor system.

60

mL of inoculum. A common inoculum was used for all ﬂasks to reduce error due to
soybean variation. Cultures were maintained on an orbital shaker at 125 rpm.

5.2.5 Analytical

Isoﬂavonoid assays based on the method of Graham (1991), and respiration
measurements of suspension and immobilized culture have been reported elsewhere
(Section 4). Sugar assays were conducted using high performance liquid chromatography
(HPLC). A Waters (Milford, MA) 600 multisolvent pump, a 25 cm by 4.6 mm Supelco
(Bellefonte, PA) LC-NHz column, and a Waters 410 differential refractometer comprised
the system. Gas chromatography of the MFB head space was performed on a Perkin
Elmer (N orwalk, CT) Autosystem Gas Chromatograph with a thermal conductivity
detector. Fixed gases were separated on a 9.1 m by 2.2 mm Hayesep DB 100/ 120 SS
column (Alltech) with helium as the mobile phase.

5.2.6 MFB Operation

The MFB biocatalyst (2 mm diameter) was l-B5 medium with 2% medium-
viscosity sodium alginate, 25% (w/w) stainless steel powder, and 30% (v/v) soybean
inoculum. The details of soybean immobilization and recovery, and control of biocatalyst
movement through the column using a magnetic valve for solids (MVS) have been
reported previously (Section 4). Some aspects of the protocol including ﬂow rates,
biocatalyst exchange intervals, and use of dissolved oxygen control were modiﬁed. The
volume of biocatalyst exchanged was metered by gradations on inlet and outlet tubing,
allowing precise control of residence time. Average biocatalyst volumes exchanged (or

“plugs”) were around 150 mL. Change in bead volume was assumed to be negligible. The

61

21 and 27 day MFB runs reported in this work employed both variable and constant
residence times, respectively. Biocatalyst was added to the latter every third day.
Biocatalyst was removed starting on the ninth day yielding three plugs of biocatalyst in
the column at all times once loaded.

The dissolved oxygen control loop was designed to maintain both dissolved
oxygen (DO) and head space C02 near target levels. A respiration rate and gas to liquid
mass transfer coefﬁcient estimated from previous work were used in simpliﬁed versions
of model equations described below. Target levels of 80% DO and 5% C02 were
arbitrarily chosen based on a measured shake ﬂask concentration of 5.5% after 2 weeks of
growth, and literature values for shake ﬂasks reaching 10% or more (Lee and Shuler,
1991). The mass balances yielded the percent oxygen required in the gas feed. Peristaltic
pump tubing diameters for air and oxygen were chosen to most closely match the desired
level of air enrichment. Rigorous control strategies have been published which maintain
oxygen and carbon dioxide at their set points regardless of respiration rate (Smith et al. ,
1990). However, the strategy used here utilized available equipment and afforded the

simplicity of a single control loop.

5.3 Model
5.3.1 Scope and General Assumptions

The MFB system was conceptually modeled as three physical phases: solid
biocatalyst, liquid medium, and gas head space. The gas phase was well mixed as its
recycle rate was much greater than the feed. The solid phase was treated as three separate

plugs, as each plug of biocatalyst in the column had a different age. Components of these

62

phases included in the model were cell mass, sucrose, glucose, oxygen, and carbon dioxide.
General assumptions made in model development are listed as follows:

1. No mixing between plugs of biocatalyst.
2. Homogeneously distributed cell mass throughout the solid (biocatalyst) phase.
3. Well mixed liquid phase.

4. Rate limiting mass transfer step at the gas-liquid interface in transport of oxygen and
carbon dioxide between the head space and plant cells.

5. Rate limiting cellular metabolism step in uptake of glucose and hydrolysis of sucrose.

Assumptions 1 and 2 were based on prior work which reported minimal solids
mixing and a uniform distribution of soybean cell aggregates throughout the biocatalyst
(Section 4). Therefore, the model does not include a radial dependence of cell
concentration within the alginate spheres. An alternative assumption would have been the
treatment of soybean and immobilization matrix as separate phases. However, the
alternative treatment did not have value-added beneﬁt as the measurements and
information needed to effectively use a two-phase (soybean and alginate matrix) model
were outside the scope of this work.

Assumption 3 requires that the dissolved oxygen is not signiﬁcantly depleted
across the length of the column. For steady-state and no liquid-solid mass transfer
resistance, the dissolved oxygen concentration (0,) is a function of vertical distance (2) as
described by Equation ( 4 )

d0, 1-8 (4)

63

where V2 is the liquid velocity in the z direction, Q is the volumetric oxygen consumption

rate in the solid phase, and s is the void ﬁ'action in the column. For a velocity of 3.5 cm/s,

a biocatalyst height of 40 cm, a respiration rate of 28 mg Ozlg cells-hr (Section 4), and an
average cell density of 3.0 g/L, the dissolved oxygen would be depleted by only 4.7%
across the length of the column for pre-saturation with air.

Assumptions 4 and 5 rely on the diffusion of dissolved substrates in the
biocatalyst not being rate limiting. This assumption was justiﬁed by both the biocatalyst
microscopy mentioned above, and Thiele modulus calculations presented in Section 4.
Thiele modulus calculations also indicate that diffusion of sugars is not rate limiting.
While sugar diffusion coefficients are somewhat lower than oxygen, bulk sugar
concentrations are orders of magnitude higher than oxygen.

5.3.2 Mathematical Development

The cell balance given in Equation ( 5 ) for each solid-phase plug assumes
exponential growth, and no loss of viability during the short residence time in the MFB.

dX.
7'“:-

(5)

The speciﬁc growth rate is described by Monod kinetics as a function of glucose. Other
authors have used Monod kinetics dependent on glucose (van Gulik et al., 1992), sucrose
(Guardiola et al., 1995), and internal glucose (Westgate et al., 1991). Applying
assumption 5 above, solid and liquid phase sugar concentrations are equal and Equation (

5 ) can be written in terms of the liquid phase glucose concentration.

64

_ GI ¢ (6)
ﬂ-ﬂmax KG+GI

 

In addition to carbon and energy source, the growth rate was also assumed to be a
function of dissolved oxygen and dissolved carbon dioxide concentrations as deﬁned by

Equation ( 7 ).

 

¢_ 0, c, (7)
- K0+0, KC+C,

Carbon dioxide was incorporated into the growth expression in order to account for
DGM, as it was readily measured and represents the major component of DGM
(Schlatrnann et al., 1997). While both the dependence (Archarnbault, 1991; Maurel and
Pareilleux, 1985) and independence (Kim et al., 1991; van Gulik etal., 1994) of growth
rates on CO; concentration have been reported, other DGM components which may
affect growth rates would be accounted for by the model.

The balance for sucrose in each of the 3 solid phases is described by Equation ( 8 )

(IS 5. (8)
bl = Qs-khyd'Xt'VbJ _'—bJ—_

 

where sucrose is consumed by extracellular hydrolysis and replenished by the liquid
medium. Sucrose is hydrolyzed by cell wall invertases (Fowler, 1982) and has been
previously described with Michaelis-Menten kinetics (Hooker and Lee, 1992). Equation (
9) is the liquid phase balance for sucrose which includes the net liquid ﬂow through the

system and loses to the biocatalyst.

65

d8 (9)
_d_tl=Fl(Sl,0 ‘S1)-Qs
Combining Equations ( 8 ) and ( 9 ) yields Equation ( 10).
d3, ash. Sb‘ (10)
7V1: FI(SI,0 —Sl)—;['—dt—J'Vbj - khdetVbJ[m

However, the assumption of negligible rate limitations for liquid-solid mass transfer
(assumption 5 above) makes all liquid and solid phase concentrations equal. This allows
rearrangement of Equation ( 10 ) to Equation ( 11 ) and elimination of the solid phase

sucrose concentration variable.

as S (11)
7;].[Vl+;Vb1)=FI(SI,O-Sl)_khyd[m];XJ/bu
The glucose balance in each solid phase is given by Equation ( 12 )
(1ij s J (12)
—V.=Q +kYX,V.———’——-q
dt b,t G hyd GS (”[Khyd'I'S’ G

which includes transfer ﬁ'om the liquid, generation from hydrolysis, and uptake by the

plant cells. Glucose is consumed for growth and maintenance as deﬁned by Equation ( 13)

I3
qG = [Y1 + %¢]X1Vbi ( )
XG

which includes a dependence for maintenance consumption on dissolved oxygen and
carbon dioxide concentrations. Equation ( 14 ) is the liquid phase balance for glucose

dG
7V1 = F1(G1,0 _Gl)_QG

(14)

66

which includes the net ﬂow of medium through the system and loss of glucose to the
biocatalyst. As with sucrose, Equations ( 13 ) and ( l4 ) can be combined into a single

balance described by Equation ( 15 ).

(15)

 

dG S
—dti[V, +2V..j= FIG... —G,)+[k..Y.. K ’ — Y” ...,¢ 20%..)
t‘ XG r'

hyd+Sl

A combined balance for oxygen, Equation( 16 ), can be written by combining the

liquid and solid phase balances as done with sucrose and glucose.

d0 0 ( 16)
—I V! + 2V0; = kL "i - 01 V1 " [Yoxu + m0¢]2(XlVb,i)
d! I H 0 i
Equation ( 16 ) includes transfer of oxygen ﬁ'om the gas phase and consumption by the
plant cells for grth and maintenance. The gas phase balance for oxygen is given in
Equation ( 17 )

d0

0 ( 17 )
dts Vg = 173(ng — 08) — kph—150- — 0, )V, + kp,0a(0, — 0g)

 

which includes the net ﬂow of gas through the system, transfer of oxygen to the liquid
phase, and transfer of oxygen due to permeability of the silicone tubing. Similar balances
for carbon dioxide are given in Equations ( l8 ) and ( l9 ).

dC Cs ( 18 )
stin+2v~1=ktirqudeceit”)

dng= -F(C Cg)-kLa[——C,]V,+kca(0— C) (19)

67

Equation ( 18 ) differs from Equation ( 16 ) in that carbon dioxide is generated rather than
consumed. The amount of dissolved C02 reaction with the liquid medium is very low
(Schlatrnann et al. , 1997) and was not included in the model.

5.3.3 Solution Method

Equations ( 5 ), ( 11 )and( 15 ) - ( 19 ) form a set of9 differential equations, as
the balance in Equation ( 5 ) is written for each of the three solid phases. The continuous
model was well suited for representation aspects including continuous feeds, exponential
cell growth, and dynamic substrate concentrations. However, incremental plug-ﬂow
movement of the solid biocatalyst represented a discontinuity. The system was therefore
solved in a piece-wise fashion. The differential equations were solved by the Fourth
Order Runge-Kutta method with adaptive stepsize control for each interval between
biocatalyst movement using Mathcad version Plus 6 for Macintosh. The results from
each interval were passed as initial conditions to the next, including step changes in cell

concentration and medium composition due to exchange of biocatalyst.

5.4 Results
5.4.1 Model Parameters

Model parameters were determined by experimental measurement, literature
sources, and ﬁtting. The list of non-adjusted model parameters and their source is given in
Table 4. The value of the gas-liquid mass transfer coefﬁcient (kLa) was determined
directly from an MFB experiment using Equation ( l7 ) assuming steady-state conditions,
as both liquid and gas phase oxygen concentrations were measured, and other parameters

known. The same mass transfer coefﬁcient was assumed for carbon dioxide balances. The

68

permeability mass transfer constants for silicone tubing were calculated ﬁom gas
permeability constants (Cole Parmer, Chicago) given as volume per time as a ﬁmction of
pressure driving force and tube thickness and surface area. The oxygen saturation constant
was determined by ﬁtting a Michaelis-Menten kinetic model to suspension-culture batch
respiration data.

The list of adjusted parameters is given in Table 5. Literature values for model
parameters, where available, were used to bound the range of adjustment. Table 5 gives
high and low literature values and the corresponding references. Optimization by
goodness of ﬁt of the model indicated that the saturation constant for hydrolysis (Khyd)
was much greater than the sucrose concentration, yielding ﬁrst-order kinetics with respect
to sucrose concentration. A ﬁrst order rate constant (k ’hyd) was deﬁned as the ratio of khyd
and KM. Values for these maximum rate and saturation parameters of 6.8 g/g cells-day
and 0.5 g/L, respectively, were given by Hooker et al. (1992). These literature values
could not predict the rapid hydrolysis seen during MFB startup. Glucose yield (Y X6) and
maintenance coefﬁcients (mg) were optimal by goodness of ﬁt at the low and high ends of
ranges, respectively, given in Table 5. This combination gives the highest glucose
consumption rate possible for the given ranges. The glucose saturation coefficient (Km) for
growth was also found to be optimal at the high end of the range, introducing a signiﬁcant
dependence of growth on glucose concentration. Literature values for the C02 saturation
coefﬁcient for growth (KC) were not available. The remaining metabolic parameters were
found to be optimal at intermediate values. Feed gas ﬂow rates (F g) as a function of time

were not available due to erroneous PC clock output, although a time-averaged value of

69

1.4 L/day was calculated. The ﬂow rate was adjusted for each interval in order to mimic

action of the dissolved oxygen control loop.

Table 4. Non-adjusted model parameters.

 

 

Parameter

Value

2.7 L
0.7 L
0.115 L/day

37.2 (g/L gas)/(g/L liquid)

156 (g/L gas)/(g/L liquid)
240 day“1

16.4 L/mz-day

41.5 L/mz-day

0.0547 m2

13.7 g/L

3.06 g/L

25%

14.9

0.564

0.4 g COz/g cell mass
0.4 g Ozlg cell mass
0.53 g glucose/g sucrose

0.38 mg/L

Source

measured by displacement
measured by displacement
measured

calculated, Perry et al. (1984)
calculated, Perry et al. (1984)
measured

calculated, Cole Partner
calculated, Cole Panner
calculated from tubing length,
diameter

measured

measured

measured

measured

measured

van Gulik et al. (1992)

van Gulik et al. (1992)
calculated, mass balance

measured, respiration data

 

 

70

Table 5. Adjusted model parameters.

 

 

Parameter Eng R_a_nge of Values
k ’hyd 1.67 Mg cells-day -
pm 0.39 day“ 0.105 - 0.468
Guardiola et al. (1995), van Gulik et al.
(1992)
Y XG 0.28 g cells/g glucose 0.28 - 0.693
Westgate et al. (1991), van Gulik et al. (1992)
K, 6.0 g glucose/L 0.5 - 6.36
Westgate et al. (1991), van Gulik et al. (1992)
KC 0.3 mg COzlL -
mG 0.204 g glucose/g cells-day 0.142 - 0.204
van Gulik et al. (1992), van Gulik et al.
(1993)
mg 0.161 g Ozlg cells-day 0.151 - 0.201
van Gulik et al. (1992)
Inc 0.222 g COz/g cells-day 0.208 - 0.277
van Gulik et al. (1992)
F8 0 - 2.2 L/day N/A

 

5.4.2 Biocatalyst Loading and Growth

Two MF B experiments were conducted which utilized gas phase recycle and
dissolved oxygen control. Cell density and speciﬁc growth rate results are shown in
Figure 17 for the second experiment. Biocatalyst loading (cell density) is reported as mass
of dry tissue per volume of biocatalyst. Data are presented for 27 days of operation, as

microbial contamination was detected on day 30. Feed biocatalyst data points which

 

71

correspond to product biocatalyst data are connected by a line for clarity. All feed plugs
had a nine day residence time. Growth data are shown for each product biocatalyst point,
and were calculated from initial and ﬁnal cell densities of that plug, assuming exponential
growth. Figure 17 shows that magnitudes in feed concentrations are reﬂected in product
concentrations. Magnitudes are particularly well matched past 10 days, where grth

rates are fairly constant. Speciﬁc growth appeared to become steady at approximately

 

 

 

 

 

 

 

 

0.15t1ay".
12 0.18
—El—Product
A101» +Feed ~r0.15
g, —6—Growth
v 8 ~ 0.12 .7"
>5
.6 it
E 61 0.09 m
‘5 iii
3 4 — --0.06 CE
C
O
O 2 — “0.03
o , , . 0.00
0 10 20 30
Time (days)

Figure 17. Biocatalyst plant-cell loadings and speciﬁc growth rates.

Similar results (not shown) were obtained for the ﬁrst experiment, although
operation lasted only 21 days. This experiment used longer and varied residence times of
11-18 days. Long residence times and growth rates up to nearly 0.12 day'1 resulted in
soybean growth which joined biocatalyst particles, preventing MV S operation. Product

biocatalyst densities reached over 11 g/L, where 10 g/L is considered the practical limit

72

before growth outside the immobilization matrix occurs. The data from this experiment
were not used for modeling they involved non-uniforrn residence times.

Model predictions for soybean concentrations in the biocatalyst are shown in
Figure 18 along with experimental data from Figure 17. Each line tracks the cell loading of
a plug as it moves through the MFB over a nine day residence time. Feed concentrations
were used as initial values. The model matched data well for biocatalyst harvested initially
(day 9) and once growth rates became steady (days 21, 24, and 27). Speciﬁc growth rates,
therefore, also matched well. The model could not accommodate the rapid increase in

grth from days 10-15, underpredicting product biocatalyst loadings on days 12 and 15.

 

 

 

 

 

 

 

 

8
7 _ :1 Product Data D
A —Modei
g 6w 0 FeedData
3’
5
c
.9
E 4
E
or 3
8
o 2
0
1
0 . i i i i t i . i .
O 3 6 912151821242730
Time (days)

Figure 18. Data and model predictions for plant-cell loading of each biocatalyst plug.

5.4.3 Sugar Hydrolysis and Consumption

Data for sucrose, glucose, and fructose concentrations and model predictions for

sucrose and glucose concentrations over the 27 day experiment are shown in Figure 19.

73

Rapid hydrolysis of sucrose to glucose and fructose was observed, resulting in sucrose
concentrations of less than 2 g/L past day 9. Glucose and fructose concentrations
increased accordingly, with glucose falling again due to consumption. Little or no
consumption of fructose appeared to take place. These results were expected as diauxic
growth of plant cells has been reported (Westgate et al., 1991), and observed previously
by the authors (unpublished data). Initial glucose and fructose concentrations were
apparently the result of hydrolysis during the autoclave cycle. Model predictions match
data reasonably well, although the model can not account for the apparent lag during the
ﬁrst 3 days of operation. Medium feed to the system included both the continuous feed
started at day 9, and the medium exchanged with each biocatalyst exchange. The medium
exchanges were step changes and give model predictions the saw tooth shape seen in
Figure 19. Sugar assays were performed at the end of each interval and correspond to the
saw tooth minima before each jump in concentration.

5.4.4 Head Space Composition and Dissolved Oxygen

Figure 20 illustrates the 02 and C02 composition of the MFB head space and
dissolved 02 (DO) concentration in the medium. Model predictions are also shown for
each variable. Head space 02 and C02 content are given as percent volume. D0 is shown
as the percent of 02 saturation, where medium in equilibrium with air would have a DO
value of 20.9. Dissolved 02 control and the feed gas pump were started on day 6 once
oxygen uptake became appreciable. Head space assays were started when the GC became
available on day 9. The model ﬁt the data well excepting days 10 and 11 when a leak in

the head space sampling line resulted in a signiﬁcant ﬂow of air into the head space. As

74

expected, the DO level remained steady at the setpoint of 16.3 (78% air saturation)
excepting the disturbance caused by biocatalyst exchange at day 21 when vacuum was not
available. The model shows a periodic behavior as a single feed gas ﬂow rate is input at
each interval, while in reality the controller matched the ﬂow rate with increasing soybean
concentration and oxygen uptake rate. Oxygen uptake rates were also calculated from the
data shown in Figure 20 using Equation ( 17 ) assuming a constant gas ﬂow rate of 1.4
L/day. Uptake held constant after day 12 at approximately 0.2 g Oz/day. The simulation
predicted an average uptake rate of 0.22 g OZ/day for the last four intervals. The average
uptake rate for the same period calculated from a previously measured immobilized

soybean respiration rate of 8.9 mg 02/hr was 0.39 g Oz/day.

 

 

 

 

 

 

 

 

15
A I o fructose D glucose;
A A sucrose Model 1
_1 a
\
3’ 10+ 9
C 8 0 0 o 0
.9 .
E i . . ..
E
8 5~ A
S
0 3
A . - - -
o. i i s A 4 a i i .
0 3 6 9 12 15 18 21 24 27
Time (days)

Figure 19. Sucrose, glucose, and fructose concentrations in MFB medium.

75

 

 

 

 

 

 

 

 

 

 

c
.2
‘9'
3
to
(D
6“
o 13 Oxygen A Carbon Dioxide
i 10 r o [1) —Model
0
s\°
Q) _
E 5
2
O
>
0 t .
O 1 O 2 O 3 0
Time (days)

Figure 20. 02 and CO; content of head space, and dissolved 02 (DO).

5.4.5 Isoﬂavonoid Production

The concentrations of daidzein and genistein in soybean recovered from product
biocatalyst are shown in Figure 21 on a dry soybean mass basis. Values are reported as
mass per mass of dry tissue. The error bars represent the 95% conﬁdence interval.
Reported values are for total isoﬂavonoids, including both free isoﬂavonoids and
isoﬂavonoids freed from moieties by acid hydrolysis. Isoﬂavonoids levels remained fairly
constant and average values were 980 and 130 ppm for daidzein and genistein,
respectively. The results in Figure 21 are an improvement over the prior 21 day MFB nm
discussed above in which concentrations averaged 180 and 30 ppm for daidzein and
genistein, respectively. The results also compare well to prior MF B work in which
average concentrations were 80 and 50 ppm (Ames and Worden, 1997). However, assays

of stock cultures used as feed in the ﬁnal MFB run averaged 24,000 and 5,700 ppm for

76

daidzein and genistein, respectively. An average of 2 stock cultures from the 21 day run
had 4,900 and 2,900 ppm. Lower yet were the averaged concentrations ﬁom several stock
cultures assayed prior to the original MF B work reported at 340 and 560 ppm for

daidzein and genistein, respectively.

 

 

 

 

 

 

{.1

a I i E I E
0 10 20 30
Time (days)

T

2000
A oDaidzein
E uGenistein
2: 1500 -~ 0
.5 5 r:
*5 1000 1-
E
or
o
1::
o
O

 

 

 

Figure 21. Isoﬂavonoid concentrations in product biocatalyst soybean.

The effect of the antibiotic/antimycotic cocktail on soybean growth and
isoﬂavonoid production was evaluated in batch shake-ﬂask cultures. Suspension culture
concentrations are shown in Figure 22. Replicate results are shown. Average soybean

speciﬁc growth rates were 0.35 i 0.05 and 0.33 i: 0.01 d'1 for control and antibiotic

cultures, respectively. The growth rates were calculated from the ﬁrst three data points,
considered to be log phase. As expected, the cocktail did not have a signiﬁcant effect on
growth rate. Daidzein and genistein concentrations on a mass of dry soybean basis are

shown in Figure 23 and Figure 24, respectively. Changes in isoﬂavonoid concentrations

77

appeared to be independent of antibiotic effects for the ﬁrst 6 days in batch culture,
decreasing ﬁ'om initial values. Past 6 days, isoﬂavonoid concentrations increased for the
control cultures, and continued decreasing for the antibiotic cultures. The change from
decreasing to increasing concentrations appears to correspond with growth phase. The
cultures no longer exhibited log grth past day 6. While temporally dependent, the
effect of the antibiotic cocktail on isoﬂavonoid concentrations is signiﬁcant. At day 9,
average daidzein and genistein concentrations were less than 30% and 20% of the control

values, respectively.

 

 

 

 

 

 

10
3 n
z o 3
.c.» 3
E 1 7
‘5 3
§ oControI (rep. 1) i
8 nControl (rep. 2)
IAAntibiotics (rep. 1)
LoAntibiotios (rep. 2)
0.1 r : % i.
0 3 6 9 12

Time (days)

Figure 22. Effect of antibiotic cocktail on soybean cell growth.

78

 

 

 

 

 

 

 

8000
7000 IOControl (rep. 1)
A |uControl (rep. 2)
S 6000 ‘AAntibiotics (rep. 1)
a- . . .
z 5000 loAnthIotIcs (rep. 2)
.9
E 4000 _.
E 3000 ~-
§ 0
8 2000 -- 8
1000i—
3 a
O 4 4 i .
0 3 6 9 12
Time (days)

Figure 23. Effect of antibiotic cocktail on daidzein concentrations in soybean.

 

 

 

 

 

 

 

 

3500
oControl (rep. 1)
3000
A DControl (rep. 2)
CE). 2500 AAntibiotics (rep. 1)
3 oAntibiotics (rep. 2)
,5 2000 ~-
6
:s 1500 5?
§
5 1000 «r 2 8
0
500 .. S
A
o
0 s 1 9 ﬂ
0 3 6 9 12

Time (days)

Figure 24. Effect of antibiotic cocktail on genistein concentrations in soybean.

Volumetric productivities were calculated for the ﬁnal MF B run and are presented
in Figure 25 along with total biocatalyst volumes contained in the MFB. Volumetric
productivity is reported in S/L bioreactor-day assuming a value of $1000 per gram of
isoﬂavonoid, and a 1 L working volume for bioreactor (column only). Shuler et al. (1990)
has noted that the primary determinant of economic viability for a bioprocess, the
volumetric productivity of the reactor times the wholesale price of the product, is 0.12

$/L-day +/- 60% for a wide variety of fermentations. Daidzein productivity falls into this

79

range, but genistein productivity does not.

0.5

0.4 ‘-

0.2

Volume (L)

0.1““

0.0

 

0.3 ~-

 

 

 

 

 

 

0.3
A 3 A 6
A A A
0 -~ 0.2
O O
.. ABiocatalyst Volume i
oDaidzein . i” 0-1
oGenistein ;
° 0 o o o
o
0 10 20 3°
Time (days)

Productivity ($lL-day)

Figure 25. Total volume of biocatalyst and volumetric productivity.

5.4.6 Modeling of Steady State Operation

The experimental data shown in Figure 17 and Figure 19 indicates steady growth
rates and sugar concentrations were achieved. Modeling presented in Figure 19 predicted

that glucose concentrations, and thus growth rates, were still changing. An extended model

80

simulation is shown in Figure 26. All feed biocatalyst concentrations were 1.3 g/L starting
at day 21. The cell density line shows the loading for the bottom plug. Cell concentrations
leaving the MFB correspond to the top of the saw tooth waves. Concentrations were
zero for the ﬁrst 6 days as biocatalyst was present only in the top and middle plug
positions. Figure 26 shows that the glucose level requires around 60 days to reach steady-
state. This is in agreement with the heuristic for time to reach steady-state of 3 to 4
residence times (F ogler, 1992), as the liquid residence time was approximately 18 days.
Sucrose levels out much quicker as the hydrolysis reaction is rapid. The glucose
concentration appears to govern the time to steady-state. As the range for Km values in
Table 5 is large, the K," value of 6 g/L may overestimate the effect of glucose concentration
on growth. Figure 27 shows a simulation for a Km of 0.5 g/L. The feed biocatalyst was 1.3
g/L for all intervals. Initial and feed medium concentrations were 20 g/L sucrose and 0 g/L
glucose. While the time for glucose levels to reach steady-state is essentially the same as
Figure 26, biocatalyst loadings and cell growth rates reach steady-state essentially within
21 days. This length of time reﬂects the biocatalyst movement rather than liquid residence
time. Model simulations predict that the time to steady state could also be reduced by
choice of sugar concentrations in the initial and feed medium. A simulation (not shown)
with a K, of 6 g/L can reach a glucose and growth rate steady-state in approximately 21
days by using an initial and feed reservoir concentration of 7 g/L glucose, and medium

containing 25 g/L glucose for biocatalyst additions.

81

 

 

 

 

 

 

 

 

1 4
sucrose
A 12 1’ /
_l
\
3 10 —-
cells

c
.g 3 ‘ glucose
9 /
E 6 1
8
c 4 -
o
O

2 a

o l : i i .

0 1 0 20 3O 40 50 60
Time (days)
Figure 26. Extended simulation of MFB run.
20
/ sucrose

3 16 -~
\
23
c a.
.2 12 cells
‘5
1: glucose
c 3 «
m
o
c
o
0 4 ~

0 . i .L + 1' +

0 1 0 2 0 30 40 5 O 6 0
Time (days)

Figure 27. Simulation of MFB using a Km of 0.5 g/L and 20 g/L sucrose feed.

82

Model simulations also provided insight into use of different biocatalyst residence
times. Figure 28 shows simulations for 9 and 12 day residence times. Both simulations
had a 1.3 g/L feed biocatalyst concentration, and feed and initial medium concentrations of
20 g/L sucrose. Sucrose levels are not shown in Figure 28 for visual clarity. Biocatalyst
loadings were higher in the 12 day case as expected, and reached over 8 g/L. The
magnitude of saw-tooth changes in glucose concentration showed no signiﬁcant increase in
the 12 day case compared to the 9 day case. The use of the longer residence time does not

create excessive periodic ﬂuctuations.

 

 

20 10
3160 :7
\ \
U’ U)
v V
C -~ C
.212 .2
T6 iii
:- 1:
C 8" C
0 Q)
0 0
C C
O O
O 4" O

 

 

 

0 10 20 30 40 50 60
Time (days)

Figure 28. Simulations of MFB for 9 and 12 day biocatalyst residence times.

5.5 Discussion
The results depicted in Figure 17 represent a signiﬁcant advance in continuous
operation of the MF B with plant cell culture. While long-term operation was previously

reported by the authors, the current work demonstrated stable operation where growth

83

rates reached a relatively steady value. Mixing of the biocatalyst during exchange
appeared to be insigniﬁcant as feed and product biocatalyst loading levels closely
paralleled each other over time. The dissolved oxygen control loop, which fed oxygen-
enriched air to the system, successfully maintained a high dissolved 02 level and head
space C02 levels in the target range. Problems which remain with MF B operation include
contamination and cell growth outside of the alginate spheres. While only the ﬁnal MF B
suffered contamination, difficulties in aseptic operation of a continuous, plant-cell
bioreactor are expected to increase with scale. Plant cell growth outside the
immobilization matrix posed a unique problem for the MFB conﬁguration. While
adhesion to reactor surfaces or signiﬁcant clogging of the base plate ﬂuidization mesh
were not observed, interbiocatalyst attachment prevented solids removal at the end of the
21 day run. Feed biocatalyst loadings and residence times must be chosen to avoid this
problem. In general, any system which has extended periods of immobilization particle
contact may be susceptible. MFB operation may beneﬁt from higher liquid velocities
which yield more open packing structures such as chain-channel (Thompson and Worden,
1997), as opposed to the essentially frozen bed regime used here.

The isoﬂavonoid production levels presented in Figure 21 also represent an
improvement over prior MF B operation. Product accumulation also showed less
ﬂuctuation over the course of the experiment. This is likely due to stable levels of
dissolved oxygen and carbon dioxide. Despite this improvement, the kinetics of
isoﬂavonoid production and degradation in the MFB are not fully understood. Lacking a

fundamental basis, mass balances for daidzein and genistein were not included in the

84

modeling. MFB beneﬁts including tight control of biocatalyst residence time, and
independent control of solid, liquid, and gas residence times, need to be linked to
isoﬂavonoid accumulation kinetics to be fully realized. MFB product levels were
signiﬁcantly below those of stock shake-ﬂask cultures. Shake-ﬂask tests determined that
the use of the antibiotic/antimycotic cocktail was partially responsible for the lower
values. Furthermore, long-term changes in soybean cell culture appear to be the most
important factor in productivity results to date. Roughly a 2 order of magnitude increase
in stock culture isoﬂavonoid concentrations has been observed over a 1.5 year period.

The MFB model yielded good prediction of process variables using metabolic
parameters within ranges available in the literature. The exception was the use of a ﬁrst
order rate constant for hydrolysis. Otherwise, use of parameters outside of the literature
bounds did not give signiﬁcant improvement in model ﬁts. A growth rate dependence on
both glucose and dissolved C02 concentrations was important to the model’s ability to ﬁt
the data. The CO; term is signiﬁcant in matching the initially very slow growth rate
followed by rapid increase before leveling off. In reality, one or more dissolved gaseous
metabolites (DGM) may be important to growth (Schlatrnarm et al., 1997). C02
monitoring provided a measure of DGM and incorporation of DGM into the model.
Concepts such as medium conditioning with endogenous growth factors may explain gaps
between data and modeling at some time points. However, the model in its presented
form relies only on variables easily measured in a typical bioprocess.

An alternative C02 term is that which includes inhibition, as used by Schlatrnann

et al. (1997) to model ajmalicine production. The effect of including an inhibition constant

85

is illustrated in Figure 29. Growth data are plotted against average dissolved C02
concentrations to which each slug was exposed as predicted by the simulation. The
relationship between C02 and growth is evident. While the inhibition model gives a better
ﬁt, the differences are small in the range of concentrations found in the MFB. The grth
rates shown were also a function of a number of other variables besides C02. Neither the
data nor the model ﬁts in Figure 29 agree with the ﬁndings of Archarnbault (1991)

regarding the gas phase where 2 and 5% C02 increased and decreased the growth rate with

respect to no C02, respectively.

 

 

 

 

 

 

 

0.2
D
5" 015 .. __..— ~~_
v /
a ,’
g 01 ,
2 I
3 :——-Model wl Inhibition]
0 0.05 D Data
Model
B
o ,L 4 1 .
0 1 2 3 4 5 6

Dissolved 002 (% sat.)

Figure 29. Comparison of saturation model with and without inhibition.

Modeling provided new insight into operational protocol for the MF B. As
biocatalyst was exchanged at intervals, along with medium required for the exchange,
periodic change in process variables was introduced. Simulations indicated that the

amplitude of the periodic change is relatively small. Based on Figure 28, operation with a

86

12 day residence time should be a good choice. Biocatalyst loadings are higher without
exceeding 10 g/L, and no signiﬁcant increase in the amplitude of glucose periodic change is
predicted. From results of 9 and 12 day residence times, it can be inferred that a 15 day
residence time would risk bioreactor failure due growth outside of the biocatalyst. The
increased soybean loading and therefore increased hydrolysis rate for the 12 day case also
results in a quicker approach to the steady-state glucose concentration. This can be
explained as the time to reach steady-state is inversely proportional to the rate for fast,
ﬁrst-order reactions in a CSTR (F ogler, 1992). Finally, the model was used to predict
medium sugar concentrations which would provide the quickest approach to steady-state.
Such strategies would require veriﬁcation, as the model’s robustness depends on the
assumption that growth rates are more sensitive to glucose concentration than other
factors. The model is also limited to conditions where glucose is the rate limiting carbon

source as it does not include diauxic growth.

5.6 Conclusions

Continuous operation of the MF B bioreactor utilizing soybean cell culture has
been demonstrated past start-up to the point of steady growth rates. A simple DO and
C02 control strategy was able to maintain these variables near target levels. An
unstructured model relying primarily on literature values for metabolic parameters
provided a reasonable description of the process. The dependence of growth on both
glucose and CO; was important to a good model ﬁt. Model results indicated that time to

achievement of steady-state is controlled by the liquid phase residence time and initial

87

conditions. Model results also predicted an optimum integer interval between biocatalyst
exchange of 4 days, and that periodic changes in the medium would not increase
signiﬁcantly over the 3 day scenario. Contamination of the long-term continuous
bioprocess and plant cell growth on the exterior of the immobilization matrix remain
concerns for process feasibility. While isoﬂavonoid productivity was improved from
earlier runs, possibly due to stable environmental conditions, concentrations were well
below current shake ﬂask levels. The use of an antibiotic/antimycotic cocktail was partly
responsible for the lower values. Research is needed to better understand the relationship

between isoﬂavonoid production kinetics and bioreactor process variables.

5.7 Nomenclature

5.7.1 Variables

[,1 speciﬁc growth rate (day'l)

0 area for mass transfer (m2)

C carbon dioxide concentration (g/L)

F volumetric ﬂow rate (L/day)

G glucose concentration (g/L)

H Henrys law coefﬁcient

K saturation coefficient (g/L)

k ’hyd hydrolysis constant (L/g cells-day)

kLa gas-liquid mass transfer coefficient (day'l)

kp tubing permeability constant (L/mz-day)

Q5
Q0

Yab

88

maintenance coefﬁcient (g/g cells-day)

oxygen concentration (g/L)

uptake rate (g/day)

rate of sucrose transfer (g/day)

rate of glucose transfer (g/day)

sucrose concentration (g/L)

volume of solid, liquid, or gas phase (L)

dry cell mass concentration in biocatalyst (g/L)

yield coefﬁcient of compound a per compound b

5.7.2 Subscripts

0

hyd

in!

max

(null) feed

ambient

biocatalyst (solid) phase
gas phase

sucrose hydrolysis
index for solid phase
interval

liquid phase

maximum

6. MFB BIOREACI' OR SCALE UP

6.1 Introduction

Previous sections have described the beneﬁts of a MFB bioreactor for plant cell
culture and demonstrated its successful operation at the laboratory scale. Additional
considerations are required, however, in designing a pilot or manufacturing scale MFB.
These considerations fall into two categories: requirement and beneﬁt of using a three-
phase system, and economics of magnetic ﬁeld generation.

MF B bioreactors can be operated with two or three phases. The third phase is a
gas which replenishes the dissolved oxygen consumed by the plant cells. While systems
described in Section 4 (Bramble et al., 1990; Ames and Worden, 1997) were two-phase,
the use of a third phase would be required at a scale where dissolved oxygen depletion
was signiﬁcant. The third phase can be supplied simply by a ring sparger as in a
traditional fermentor. The physical characteristics of such a three-phase system have been
studied in the lab of Dr. Mark Worden (Thompson and Worden, 1997). Alternatively, the
third phase can be supplied by microbubbles (described in Section 3.3). Microbubbles not
only provide high gas-liquid mass transfer rates, but may also reduce operating costs of a
three-phase MF B. Due to their size, microbubbles cause less disruption to the bed than
traditional sparger bubbles. Solenoid power, and thus cost, is directly related to the
amount of disruption which must be overcome in the bed to prevent solids mixing.

For a MFB to be feasible at large scale, generation of a suitable strength magnetic

ﬁeld must be practical. Impractical situations would include excessively large solenoid

89

90

equipment or dramatic increases in power consumption with size. The analysis provided
here compares MFB scale up with that of a traditional CSTR ferrnenter and illustrates the
feasibility of the MFB. Additional capital costs of the MFB over the CSTR are also
considered. While any of several parameters could be kept constant for scale-up,
geometric similarity was chosen. Two cases of geometric-similarity scale up were studied.
The ﬁrst stemmed from an inherent optimum in the design of magnetic coils. There exists
an optimum geometry where the power consumption, regardless of other considerations,
will be minimized. With choice of reactor aspect ratio, an optimum coil thickness will
exist which minimizes power consumption. The drawback to using optimum geometry is
that coil thickness can become quite thick, and thus expensive. The second case was
therefore and economic analysis considering operating costs versus capital cost of the
coils. An optimal coil thickness was determined by minimizing the present value cost of

the project.

6.2 Materials and Methods

Soybean culture and respiration measurements were described in Section 4.
Equipment for microbubble studies including a 80 cm by 5 cm inner diameter magnetized
ﬂuidized bed system and 4 mm diameter alginate beads with 50% magnetite by weight
have been detailed by Thompson and Worden (1997). Reverse-osmosis water was
metered into the column and distributed through a 24 cm calming region. Compressed air
was metered in just above the calming region through a stainless steel ring sparger 3.5 cm

in diameter with 0.2 mm holes spaced at 5 mm intervals around the ring. Microbubbles

91

were metered into the liquid inlet line immediately prior to calming region. The
microbubble generator and its operation have been described by Bredwell et al. (1995a).
The unit consisted of a 6 L fermentation vessel with the impeller replaced by a 5 cm
diameter, stainless steel disk. A motor mounted above the head plate rotated the shaft at
speeds up to 9000 rpm. Baﬁles were approximately 2 mm from the spinning disk.

The power required to prevent mixing in a MFB was compared for three cases:
liquid only, liquid sparged with air ﬁom ﬁ'itted glass at the bottom of the column, and
liquid with supplemented with microbubbles. Liquid ﬂow rates were 6.1, 7.6, and 9.1
L/min, and gas ﬂow rates were 0.24 and 0.54 L/min. For microbubble aeration, the gas
fraction of the foam (approximately two-thirds) was used to determine the ﬂow rate
equivalent to 0.24 or 0.54 L gas/min For each combination of ﬂow rates, the solenoid
current (and power) was increased ﬁ'om zero in 0.2 A increments, observing the bed
through upper and lower view ports for 20 to 30 seconds. Bed regimes similar those
described by Thompson and Worden (1997) were observed at each increment, while
looking for the point at which mixing was prevented. Power was calculated from current

and voltage meters on the power supply.

6.3 Theoretical
6.3.1 Oxygen Depletion in the Liquid Phase

A simple model neglecting mass-transfer resistances where a pre-saturated liquid
phase ﬂows up through a solid phase containing soybean has been used. For steady-state,

plug-ﬂow liquid movement, and no mass transfer resistance between the phases, the

92

dissolved oxygen concentration (C) is a function of vertical distance (2) as described by
Equation ( 20 )

dC 1-8 (20)

where V2 is the liquid velocity in the z direction, Q is the volumetric oxygen consumption

rate in the solid phase, and e is the void fraction in the column. Integration yields

__1.:_€._Q_.h (21)

for a column of height (h) and air saturated liquid entering at the bottom. For operation at
1.6 L/min (Section 4) and an assumed void fraction of 0.4, v2 was 3.5 cm/s. The bulk

oxygen concentration at 25°C was taken to be 8.43 mg/L (Bailey and Ollis, 1986).

Respiration rates and ranges for soybean densities in the biocatalyst were taken from
Section 4. The oxygen consumption rate (Q) was calculated from the measured
suspension culture respiration rate of 28 mg Oz/g cells-hr, and an integral average cell
density of 3.0 g/L. The average density was based on feed and product biocatalyst
densities of 0.4 and 10 g/L respectively.

6.3.2 Solenoid Equations

For a uniform-current-density solenoid the power consumption is described by
Equation ( 22 ) taken from Montgomery (1969)

W_}L])é (22)

H0 =G(a,ﬁ)-[pal

93

where H, is the magnetic ﬁeld vector, G(a, B) is the F abry factor, W is power, A is the
fraction of coil cross section occupied by wire, p is the resistivity of the wire, and a ,, is

the inner radius of the coil. The Fabry factor is given by Equation ( 23 ), also from

Montgomery (1969)

 

a2—1

)%.lna+(a2+ﬁ2))é (23)

1 22:5
G .1. =-.(_
( ) 5 ”(”132)”

where a is the ratio of the outer to inner coil radii and )3 is the ratio of the coil height to
the inner coil diameter. It is notable that ﬂ is also assumed to be the aspect ratio of the
MFB. The Fabry factor can be thought of as a type of efﬁciency, and has a maximum
near a = 3 and ﬁ = 2. As an aspect ratio of 2 is reasonable for a large ﬂuidized bed, these

values were chosen for the ﬁrst scale-up method of minimum power consumption (MP).
Another important result is obtained ﬁ'om Equation ( 23 ). For a geometric similarity
scale-up (constant Fabry factor) and given magnetic ﬁeld strength, the power

consumption varies with bed radius. This indicates that

power cc volume}; ( 24 )
___power cc volume}? ( 25 )
volume

which demonstrates a decrease in power consumption per volume with increasing size.
The resistivity (p) of copper wire at the cooling water temperature of 10 degrees

Celsius was calculated to be 1.63 x 10'8 ohm-m. The fraction of space occupied by wire

94

(A) for a staggered packing was calculated to be 0.8976, Two magnetic ﬁeld strengths

were speciﬁed, 100 and 300 Gauss. A ﬁeld of 100 Gauss represents an adequate value for
operation of a MFB bioreactor using 50% magnetite, alginate beads with an approximate
diameter of 4 mm, and microbubble aeration as determined in bed studies above. A ﬁeld of
300 Gauss represents a value more adequate for operation of the same system with
conventional sparging aeration.

The speciﬁcation for magnetic ﬁeld strength is given at the axial and radial mid-
point of the MFB bioreactor. The ﬁeld will vary however, along the solenoid axis unless
the solenoid is inﬁnitely long. Solenoids with low aspect ratios (determined by the bed
aspect ratio) may have an unacceptable decrease in ﬁeld strength away from the axial mid-
point. Equation ( 26 ) from Montgomery (1969) describes the variation in ﬁeld strength

along the axis

HIE] = 1r:!.(0)[F(a”3 + 12:2?” " z/a.) ( 26 )

where Hz is the ﬁeld strength as a function of dimensionless axial position, Hz(0) is the
ﬁeld strength at the axial mid-point (maximum ﬁeld strength), and z is the axial distance

from the center of the solenoid. The shape factor (F) is given by Equation ( 27 ).

M’shmhzwq/z (27)

F(a,ﬁ)= 10 1+(1+ﬁ2))§

 

 

95
Equation ( 27 ) is used to determine the change in ﬁeld strength along the axis for the
original chosen aspect ratio ()3 = 2), as well as for an aspect ratio which yields a more
uniform ﬁeld strength ([3 = 8).

6.3.3 Economic Analysis

A present value cost analysis was performed in order to determine an minimum

cost (MC) value for a as opposed to a minimum power (MP) value given by a = 3. The
MC case was examined as very thick coils (a = 3) will yield very low power
consumption, but high capital costs for the coils. For a given volume, a was optimized to

give a minimum present value cost. The parameters used in the analysis are given in Table
6. The rate of inﬂation on operating costs was assumed to be zero as the increase in

electricity costs are expected to match the rate of inﬂation.

Table 6. Parameters for present value cost analysis.

 

Parameter we

Cost of electricity 0. 1 $/kW-hr

Cost of copper magnet wire 3.0 Mb.

Project life 10 years

Scrap value of copper coils 20% of original cost
Rate of inﬂation on operating costs 0%

Rate of inﬂation on value of scrap 40%

 

 

 

96

6.3.4 Comparison with CSTR Fermentors

A variety of methods exist for scale-up of CSTR fermenters (Hubbard, 1987)
which give very different relationships between the power per volume for agitation and
volume of the vessel. A correlation based on industrial data was therefore chosen. Einsele
(1976) correlated power input data for 0.5 to 300 m3 fermenters with Equation ( 28 ).

power ( 28 )

 

oc volume‘o‘37
volume

In order to develop a curve for comparison with the MF B bioreactor, a proportionality
constant is needed for Equation ( 28 ). Power consumption per volume for plant scale
fermenters has been given as 1 - 3 kW/m3 (Aiba et al., 1973) and 1 - 2 kW/m3 as a low
value (van't Riet, 1983). A proportionality constant was determined in order to yield 1
kW/m3 for a volume of 300 m3. The correlation correctly yields 1 -3 kW/m3 values over
the range of industrial size fermenters.

Another consideration is the additional capital cost of the solenoid for a MFB
bioreactor compared with the cost of a C STR fermenter. Comparisons were made using
fermenter cost correlations from Kalk et al. (1986) based on 1985 prices. The solenoid
cost only accounts for the purchase of copper magnet wire, but use of 1985 fermenters
costs likely balances the neglect of labor and power supply costs for the solenoid. These
neglected costs are also likely to be offset as agitator and motor costs are eliminated in a

MFB.

97

6.4 Results and Discussion

6.4.1 Oxygen Depletion in the Liquid Phase

For the lab scale MFB described in Section 4, the depletion of dissolved oxygen
across the 0.4 m height of biocatalyst is predicted to be 4.7%. Thus presaturation appears
to be satisfactory for the lab scale experimental conditions. Equation ( 21 ) yields roughly
a 1 mg/L drop in dissolved oxygen concentration for every m of height. Thus, the liquid
would be nearly 50% depleted at 4 m, and essentially devoid of oxygen at 8 m.
Maintaining dissolved oxygen at 80% of air saturation (as in Section 5) or higher
throughout the MFB would limit the height to 1.7 m. Addition of a gas phase in the MFB
would prevent both low dissolved oxygen levels and large variations in dissolved oxygen
across the column.

6.4.2 Reduction in Power Consumption by Microbubbles

Power required to prevent solids mixing was less for the microbubble case than the
sparged case at all ﬂow rate combinations. In order to quantify error in measurements,
repeated studies of the 7.6 L/rnin and 0.24 L/min case were done. The results comparing
no air, sparged, and microbubble cases for the top view-port are shown in Figure 30. Error
bars show the 95% conﬁdence interval. The reduction in power required for the magnetic
ﬁeld while supplying the same volume of air is quite substantial. This result indicates that
disturbances from rising bubbles are large compared to disturbances caused by liquid
turbulence. The disturbances may, in part, be liquid turbulence created by the bubbles.
The results for the bottom view port are shown in Figure 31. Power requirements are

much less at the bottom port, but the relationship between the three cases is similar. The

98

120

 

100 -- "'
80 -- ~——~

 

60 ~~ .-

Power (Watts)

40 -.

r11 r11

No Air Sparged Microbubbles

 

 

 

 

 

Figure 30. Power consumption required to prevent mixing at top view port.

 

 

 

 

 

14 ._
121-
1310 .___
1%
E 8 -L
'5 6T
§ 4-.
2-.
0 | E I r—I—r

 

 

No Air Sparged Microbubbles

Figure 31. Power consumption required to prevent mixing at bottom view port.

99

results also indicate that the use of microbubbles may actually decrease the power
required.
6.4.3 Uniformity of Magnetic Field

A limitation of short solenoids (and thus short MFBs) is the increasing axial
nonuniformity of the ﬁeld strength as the length decreases. The normalized magnetic ﬁeld

strength over one-half of the dimensionless length of the solenoid is shown in Figure 32
for [i = 2, and in Figtu'e 33 for 6 = 8. Two curves are plotted in each ﬁgure, one for the
MP (0: = 3) case and one for a= 1.04 which is typical for the MC case for a 12 - 13 m3

MFB. Regardless of coil thickness, the results show that aspect ratios higher than 2 may
be desirable in order to maintain ﬁeld uniformity over the entire column. With this result,

power per volume for scale-up was examined for 0 = 8.

 

0.8 _.

0.6 --

      

 

0.4 —~

 

—e—a|pha=1 .05
0.2 —~ +a|pha=3.0

 

 

 

Dimensionless ﬁeld strength

 

l

0 1 1 4. .L . . .
0 0.2 0.4 0.6 0.8 1 1.2 1.4
Dimensionless height

Figure 32. Field strength as a function of column height for an aspect ratio of 2.

100

 

0.8 --

0.6 -1

   

0.4 --

 

-O—a|pha=1.04
0.2 —- +alpha=3.0

 

 

 

 

Dimensionless ﬁeld strength

 

0 1 1 1 1 1 . .
0 0.2 0.4 0.6 0.8 1 1.2 1.4

 

Dimensionless height

Figure 33. Field strength as a function of column height for an aspect ratio of 8.

6.4.4 Power Per Volume Ratios

The results of power per volume scale-up calculations are a comparison of three
cases: (i) a CSTR fermenter, (ii) a MFB designed for minimum power consumption, and
(iii) a MFB designed for the minimmn cost. The results for the 100 and 300 Gauss
speciﬁcations and an aspect ratio of 8 are shown in Figure 34 and Figure 35. The log-log
plots show favorable scale up in all three cases. More importantly, both cases for the
MFB show signiﬁcantly less power consumption per volume than the CSTR. As
expected, the MP case sets a lower bound while the MC case falls in between the other
two. Depending on the volume and magnetic ﬁeld speciﬁcation, the MP case is roughly 1
to 3 orders of magnitude lower than the CSTR The MC case is roughly 1.0 to 1.5 orders

of magnitude lower than the CSTR.

101

 

 

 

 

1000
+CSTR
100 -» -B—MFB: Min. Power
+MFB: Min. Cost
,.. 10 “I N
"E
\ 1 -1
E o
0.01 1-
0.001 ..
0.0001 1 1 1
0.1 1 10 100 1000

Volume (m3)

Figure 34. Power consumption of 100 Gauss MFB vs. CSTR.

 

 

 

 

1000
+CSTR
100 “ —a—MFB: Min. Power
__ +MFB: Min. Cost
«7‘ 10 N
E
\ 1 _.
E 01 M
5 ° 1
0.01 1*
0.001 ~~
0.0001 1 1 1
0.1 1 10 100 1000
Volume (m3)

Figure 35. Power consumption of 300 Gauss MFB vs. CSTR.

102

The difference between the MFB cases in Figure 34 and Figure 35 is due solely to

the difference in coil thickness described by a. Clearly, the difference is quite large. An

example can be made with a column 1 m in diameter with a 100 Gauss ﬁeld. The coil

thickness for the MP case would be 1 m (a = 3). In contrast, the coil thickness for the

MC case would be 1.3 cm. This indicates that the cost of the solenoid is signiﬁcant in
comparison with the cost of power. Similarly, the coil thickness for the 300 Gauss case is
3.8 cm. The increased thickness for higher ﬁeld strength speciﬁcation is as expected. Also
of interest is the insensitivity of coil thickness to scale-up. For the range of volumes used
in calculations, coil thickness was nearly constant, increasing only slightly with volume.
As an example, the thickness for a 261 m3, 300 Gauss system is 3.9 cm.

6.4.5 Capital Costs

A MFB will have a capital cost of the solenoid in addition to the cost of the
fermenter. For the MFB to be feasible the added capital cost can not be excessive. Costs
of the solenoid were calculated as the percent of the total capital investment (fermentation
system and solenoid). Values range from roughly 2 to 10% for the 100 Gauss case, and
roughly 5 to 25% for the 300 Gauss case. Results vary with size and also depend on
whether the system is skid mounted or ﬁeld erected (Kalk et al., 1986).

6.4.6 Power Beneﬁts of Microbubble Aeration

As previously described, 100 and 300 Gauss ﬁeld strengths represent values
appropriate for microbubble and traditional sparging operation of a MFB bioreactor
respectively. Regardless of exact values, microbubble aeration has been demonstrated to

reduce the magnetic ﬁeld required (and thus power) to prevent mixing in a MFB as

103

compared to traditional sparging aeration. The comparison of power consumption per
volume for the 100 and 300 Gauss cases demonstrates the power and cost savings
potential of microbubble aeration. The 100 Gauss case has power per volume ratios
roughly one-half order of magnitude lower than the 300 Gauss case.

While additional power is required to generate the microbubbles, the power
required should be small compared to the amount saved over standard sparging. A
microbubble MFB system may consist of a MFB and a smaller CSTR for microbubble
generation. The rrricrobubble CSTR is expected to be small in comparison with the volume
of the MF B or a CSTR fermenter. The microbubble generation power is expected to be
small in comparison with the power savings incurred by use of microbubbles instead of
traditional sparging. Improved microbubble generator designs such as that of Michelsen et
al. (1990) may further reduce power requirements. An additional beneﬁt of microbubble
aeration over traditional sparging is the reduction in capital costs. A decrease in required

magnetic ﬁeld strength yields a decrease in solenoid thickness and therefore cost.

6.5 Conclusions

The MP B bioreactor is a promising reaction system for plug-ﬂow processing of
immobilized plant cells. A three-phase system is not needed at the laboratory scale, but
would become necessary at large scale. The MFB scales up favorably with regard to
power per voltune requirements. Calculations indicate that an order of magnitude
reduction in power consumption can be achieved over a traditional CSTR fermenter.

Added capital costs for the solenoid are expected to be very reasonable. MFB bioreactors

104

can be designed with aspect ratios approximating an axially uniform magnetic ﬁeld. The
beneﬁts of microbubble aeration over sparging aeration include a signiﬁcant reduction in

power required to prevent solids mixing in the MFB.

7. PLANT STUDIES IN THE DGC

7.1 Introduction

A wide variety of plant secondary metabolites have value as pharmaceuticals,
food colors, ﬂavors and fragrances. Plant pharmaceuticals include taxol, genistein, codeine,
morphine, quinine, shikonin, ajmalacine, and serpentine. Food product examples are
anthocyanins, saffron, vanilla, mint, and a variety of fruit and vegetable ﬂavors. Industrial
application of plant cell culture has been limited by drawbacks including slow growth
rates, low product yields, and poorly understood metabolic regulation. Plant growth rates
and levels of secondary metabolites produced are strongly inﬂuenced by key effector
variables such as carbon source, nitrogen source, hormones, precursor compounds and
pH. However, efforts to increase production levels typically rely on trial-and-error
studies. Because of the large number of potential effector combinations, and slow plant
cell growth rates, screening studies in shake ﬂasks or petri dishes can be very labor
intensive and time consuming. The DGC has been proposed as a tool for greatly
accelerating these trial-and-error studies.

The overall goal of this work was to establish the DGC as a tool for studying
plant cell growth and product formation as a frmction of chemical concentration. Research
was also conducted in order to demonstrate the use of image acquisition and processing as
a non-invasive assay for plant cell growth. Immobilized soybean cells were exposed to
pseudo-steady-state chemical gradients in a DGC. Images documenting cell growth in the

DGC were acquired over two to three weeks, and the cell and chemical gradients were

105

106

measured at the end of the experiment. Soybean cells were also exposed to chemical
gradients of compounds which were expected to effect secondary metabolism, and
therefore the accumulation of daidzein and genistein. Experimental protocols developed in
this study included a method of pouring DGC gels which provided a rapid approach to a

steady-state gradient, and correlation of image grayscale with soybean cell concentration.

7.2 Materials and Methods
7.2.1 Chemicals

All chemicals for Gamborg B5 medium were purchased from Sigma Chemical (St.
Louis, MO) except sucrose from J. T. Baker (Phillipsburg, NJ). Agar and 2,4-
dichlorophenoxyacetic acid were purchased from Sigma Chemical. Ampicillin was
purchased from Sigma Chemical, and Timentin was purchased from SmithKline Beecharn
(Philadelphia, PA). Helium was purchased from AGA Gas (Cleveland, OH). Acetonitrile
and acetic acid were purchased ﬁ'om EM Science (Gibbstown, NJ). All other chemicals
were purchased from J. T. Baker.

7.2.2 Diffusion Gradient Chamber System

A detailed description of the DGC has been given by Emerson et a1. (1994). The
system is comprised of a DGC (shown in Figure 36), lightbox, reservoir ﬂasks, tubing,
and peristaltic pump. The chamber, which is machined from a polycarbonate block by
Koh Development (Ann Arbor, MI), consists of an arena and recessed reservoirs. The
reservoirs are also made from polycarbonate, and each contains a stainless steel inlet port

(1.0 mm id.) and outlet port (2.5 mm id). The reservoirs are separated from the arena by

a 0.05 pm pore size membrane, allowing permeation of small molecules. The arena is 5 cm

107

by 5 cm by 2.8 cm high. The reservoirs have a volume of 3 mL. Reservoir concentrations
were maintained by pumping solutions ﬁ'om l L Erlenmeyer ﬂasks to the reservoirs
through 0.8 mm ID Norprene tubing (Cole Parmer, Chicago) using a peristaltic pump
(Cole Parmer). The solution exiting the reservoir returned to the reservoir ﬂask via 1.6 mm
ID Norprene tubing. The entire DGC assembly was sterilized by autoclaving and

subsequent manipulations were done in a laminar ﬂow hood.

 

 

 

L1
\2
a.

[)1

1 '1
l--------J§-------

 

 

 

 

 

 

 

:----§---~

[r
/

 

 

 

 

Figure 36. Diagram of the Diffusion Gradient Chamber top view.

The gel in the DGC was poured in three layers as shown in Figure 37. After tilting
the chamber, Layer 1 was poured. This layer consisted of 20 mL of 0.3% agar and the
chemoeffector of interest at the source ﬂask concentration. The angle of the chamber was
adjusted so that the gel extended along the bottom of the arena just reaching the side wall.
Once the gel had cooled and solidiﬁed, the chamber was returned to level and Layer 2 was
poured. This layer consisted of 20 mL of 0.3% agar containing the chemoeffector of

interest at the sink ﬂask concentration. Plant cell inoculum was prepared by passing a 2

108

week-old culture through a 400 micron mesh. Layer 3 was prepared by diluting a portion
of the ﬁltrate 1:2 with 0.6% agar containing the chemoeffector of interest at the source

ﬂask concentration. Once Layer 2 had solidiﬁed, 10 mL of the inoculum-agar (Layer 3)

was poured and allowed to solidify.

 

 

i 1 Sun I.

0.4 cm Layer3 (lnoculum) I

 

 

 

z
t T Layer2 2.2 cm
Y

Layer 1 i

Figure 37 . Schematic side-view of DGC showing three slabs.

 

A slightly different procedure was used for isoﬂavonoid accumulation studies.
Soybean tissue which passed a 1 mm mesh was collected in a 100 micron mesh. The cells
were added directly to 0.3% agar with the appropriate chemoeffector concentration. This
method provided greater cell mass for isoﬂavonoid extraction and quantiﬁcation.
Experiments were run for approximately 1 week, compared to 2 to 3 weeks for grth
studies.

Once the chamber lid was fastened, the DGC was placed on a lightbox that
accommodated three chambers. The lightbox had (2) 30 cm ﬂuorescent bulbs mounted on
opposing sides of the box and black felt on the bottom. This anangement provided
diffuse, dark-ﬁeld illumination of cells. The bottom of the lightbox was covered with black
felt to provide contrast as viewed from above. Images of the chambers were captured with
either a Macintosh system consisting of a Power Macintosh 7200/90 (Apple Computer,

Cupertino, CA) and a Color QuickCam (Connectix, San Mateo, CA), or a PC system

109

consisting of a Gateway P5-60 (North Sioux City, SD), WinVisionPro video capture card
(Quanta, Mountain View, CA), and a Pulnix TM-7CN CCD camera (Sunnyvale, CA).

The lightbox and cameras were housed in a temperature controlled room at 28°C.

Swelling or drying of the gel in the arena was found to be sensitive to the height
difference between the top of the agar in the DGC and the outlet of the reservoir ﬂask
return tubing. As the correct height difference could not be predicted a priori, ﬂask
heights were adjusted accordingly to prevent swelling or drying.

7.2.3 Chromatography

Sucrose and ampicillin analyses were conducted using high performance liquid
chromatography (HPLC). A Waters (Milford, MA) 600 multisolvent pump, Waters 490
multiwavelength UV absorbance detector, and Waters 410 differential refractometer
comprised the system. A 25 cm by 4.6 mm Supelco (Bellefonte, PA) LC-NH2 colmnn
was used for sucrose. A 25 cm by 4.6 mm Rainin (Wobum, MA) Microsorb-MV column

containing 5 urn C-18 packing was used for ampicillin. Gas chromatography of the DGC

head space was performed on a Perkin Elmer (N orwalk, CT) Autosystem Gas
Chromatograph with a thermal conductivity detector. Fixed gases were separated on a 9.1
m by 2.2 mm Hayesep DB 100/120 SS column (Alltech) with helium as the mobile phase.
HPLC measurement of daidzein and genistein has been described in Section 4.2.2.2.

7.2.4 Soybean Culture

Suspension cultures of Glycine max cv. Mandarin have been described in Section

4.2.5.

110

7.2.5 Grayscale and Concentration Correlation Experiments

Experiments were conducted to verify a linear relationship between light intensity
(as grayscale values) and plant cell concentration (on a dry weight basis) in the DGC. A
DGC was ﬁlled with plant cell culture and water to a total volume of 10 mL. The ratio of
culture to water was varied to control cell concentration. Various plant morphologies were

tested, including cultures that were urrsieved, cultures passed through a 400 um sieve,

cultures passed through a 1 mm sieve, and cultures previously immobilized and grown in
agar. The mean grayscale value for the DGC arena in each image was obtained using NIH
Image (National Institutes of Health, http://rsb.info.nih.gov/NIH-IMAGE/).

7.2.6 Soybean Growth Experiments

The effects of several different types of chemoeffectors were tested: carbon
source, nutrient salt, antibiotic, nitrogen source, hormone, vitamin, and isoﬂavonoid
precursor. The chemoeffectors and their source ﬂask concentrations are listed in
Image Acquisition and Processing

Images of the DGC arena were acquired over the course of each experiment with a
CCD camera. Images were acquired at intervals from 1 to 3 days at 640 by 480 pixels
resolution. In both image acquisition systems, variation in picture brightness (ﬂickering)
was observed as the image was continually refreshed. Triplicate images were taken and
averaged to reduce the effect of this variation.

Table 7. In all cases, the sink ﬂask contained none of the chemoeffector. For ampicillin

and Timentin experiments, the sterile ﬁltered antibiotics were added after autoclaving. For

111

the ammonium sulfate experiment, components of the B-5 basal salt mixture were added
individually excluding ammonium sulfate to prepare the medium.

7.2.7 Image Acquisition and Processing

Images of the DGC arena were acquired over the course of each experiment with a
CCD camera. Images were acquired at intervals ﬁ'om 1 to 3 days at 640 by 480 pixels
resolution. In both image acquisition systems, variation in picture brightness (ﬂickering)
was observed as the image was continually refreshed. Triplicate images were taken and
averaged to reduce the effect of this variation.

Table 7. List of chemical gradients used in DGC experiments.

 

 

 

 

 

Gradient Chemical(s) Source Flask Concentration Times Normal Value
Sucrose 20 and 30 g/L 1 and 1.5

Gamborg B5 Salts 9.3 g/L 3

Ampicillin 300 and 600 ppm -

Timentin 500 ppm -

Ammonium Sulfate 0.13, 0.40, and 0.67 g/L 1, 3, and 5
2,4-Dichlorophenoxyacetic 2 and 5 ppm 2 and 5

acid (2,4-D)

Thiamine 10 ppm 1

Phenylalanine 0.1 and 1.0 g/L -

 

Grayscale data were obtained over speciﬁed areas of the DGC with the plot

proﬁle function of NIH Image. Brieﬂy, a proﬁle window was chosen with dimensions

 

112

ranging ﬁom 4.2 to 4.6 cm along the gradient, and 3.3 to 3.8 cm perpendicular to the
gradient. Proﬁle window dimensions were chosen to exclude shadows from edges and
concentration gradients due to the reservoir windows not spanning the entire width of the
DGC. The plot proﬁle ﬁrnction returned an average grayscale value between 0 and 255 for
each pixel row perpendicular to the gradient. The initial image of each experiment was
subtracted from the images at subsequent times to compensate for ﬂaws in the DGC lid
and non-uniform lighting.

7.2.8 Analytical

7.2.8.1 Post-experiment Sampling of Gel

Total sugar and/or cell concentration gradients were measured at the end of
experiments. Four glass microscope slides were inserted into the gel perpendicular to the
gradient dividing the slab into 5 equally sized sections. For total sugar and ampicillin,
plugs of the entire depth of the agar were taken using a cut-off 1 mL autopipetor tip
(approximately 5 mm diameter). Samples were taken along the gradient in two lines
located 1 cm on either side of the centerline of the DGC, and centered between dividers.
Gel plugs were stored frozen until assayed. The DGC gel was then partially frozen to
facilitate removal of the individual sections.

7.2.8.2 Sugar Assays

Gel plugs in closed rnicrofuge tubes were melted in a boiling water bath. Samples
were centrifuged to settle any soybean tissue. Liquid samples from the supernatant were
then assayed for total sugar concentration by the phenol-sulfuric method (Dubois et al. ,

1956). Gel samples without sucrose were assayed to correct for agar interference in the

113

assay. Sucrose concentrations in the reservoir ﬂasks were measured by HPLC. The
isocratic mobile phase was 75% acetonitrile, 25% water, and 0.5% acetic acid ﬂowing at
1.5 mL/min. Sucrose was detected at 6.8 minutes by a differential refractometer.

7.2.8.3 Ampicillin Assays

Gel plugs in microcentrifuge tubes were melted and centrifuged as for total sugar
assays. Ampicillin stability dining heating was veriﬁed experimentally. Microcentrifuge
tubes containing gel with ampicillin were placed in a boiling water bath for 20 minutes.
Concentrations as measured by HPLC at 5 minutes intervals did not decrease. A 0.200
mL volume of the melted gel was combined with 0.600 mL of 60% aqueous acetonitrile in
a new microcentrifuge tube. The tube was vortex mixed, and injected into the HPLC.
Ampicillin was quantiﬁed with the UV absorbance detector. The isocratic mobile phase
was 25% acetonitrile, 75% water, and 0.5% acetic acid ﬂowing at 1.0 mL/min. The
ampicillin retention time was 9.5 minutes.

7.2.8.4 Dry Weight Assays of Gel Sections.

Gel sections were supplemented with approximately ten volrunes of distilled
water and heated on a hot plate to melt the agar. Water addition ensured the agar would
not solidify upon cooling. Gel plugs used for total sugar assay were returned to their
respective gel sections to account for soybean tissue in the plugs. Cell mass was collected

on dried and tared Whatrnan 41 ﬁlter paper. Samples were dried overnight at 70°C and

weighed.

114

7.2.8.5 Isoﬂavonoid Assays of Gel Sections

Gel sections were prepared as described above for dry weight assays except for
the drying step. Cell mass which had been collected on ﬁlter paper was placed in tared
microcentrifuge tubes. Extraction of cell tissue was as described in Section 4.2.8.3. Initial
experiment samples, however, were extracted once with 80% aqueous ethanol and injected
directly into the HPLC after centriﬁrgation. Hydrolysis of samples was performed once it

had been established that isoﬂavonoids were present in measurable quantities.

7.3 Theoretical
7.3.1 Unsteady-State Model of the DGC System

To study the effect of chemoeffector concentrations on plant cell growth, it is
desirable to have a steady-state concentration gradient across the chamber. During the
start-up phase while diffusion establishes the gradient, the largest variation from steady-
state occurs. The system eventually reaches a pseudo-steady-state where reservoir ﬂask
concentrations change slowly due net diffusion across the chamber. It is desirable to
minimize the time required for the pseudo-steady-state gradient to be established.
Unsteady-state modeling of diffusion in the DGC provided a method for determining the
time required for the 3-slab pouring method depicted in Figure 37 to approach steady
state. Sucrose was chosen for the modeling studies as it was the focus of initial
experiments.

Unsteady-state diffusion of sucrose through the gel is described by

'39? = DSVZS

( 29)

115

where S is the sucrose concentration and D3 is the diffusivity of sucrose. For the two-

dimensional system depicted in Figure 37, Equation 1 becomes

as 825 325 (30)
97:01???)

where y is the direction along the gradient and z is the vertical direction. The boundary

conditions at the source and sink reservoir membranes are given by

 

8S ks ( 31)
= — — S .
QTLO Ds (S|y=0 Sink)
for the sink reservoir and
— SI = _ S OUFCC
ay F5 D5 ( y 5 S )

for the source reservoir, where ks is the membrane mass transfer coefﬁcient. The source
and sink ﬂask concentrations are assumed to be constant throughout the experiment. This
assumption was validated by the data. For all other boundaries, the no-ﬂux boundary
condition is applied

g-DSVS =0 (33)

where g is the unit normal vector. The source ﬂask concentration (55mm) was set at either
20 or 30 g/L. The sink ﬂask concentration (SSW) was set at 0 g/L. The initial
concentrations in the arena gel were chosen to be either those of the three-slab pouring
method or a single slab initially containing no sucrose. The initial concentrations for the
three-slab method were 20, 0, and 10 g/L for layers 1, 2, and 3, respectively. The initial

concentrations for the single slab method were 0 g/L for all three layers.

116

Equations 2 through 5 were solved using the Partial Differential Equation (PDE)
Toolbox in Matlab 4.20 (The Mathworks, Natick MA). The PDE Toolbox uses the ﬁnite
element method with triangular elements. The PDE Toolbox was also used to determine
the steady-state solution of the model using elliptic (no time dependence) equations. The
steady-state solution was compared to the unsteady-state solution to determine the time
necessary for the depth-averaged sucrose concentration to reach 95% of the steady-state
concentration at each y-position in the chamber.

7.3.2 Parameter Determination

A diffusion coefficient (D3) of 0.0183 cmz/hr was calculated by a correlation
(Cussler, 1984). The mass transfer coefﬁcient (kg) was measured experimentally. A DGC
system was set up having one reservoir separated from the arena by a semi-permeable
membrane, and the other three sealed-off. The membrane was coated with an
approximately 1 mm thick layer of 0.3% agar to duplicate any agar effects on the
membrane used in a DGC experiment. A sucrose solution was pumped through the
reservoir from a ﬂask containing 3 L of 30 g/L sucrose and recycled as in growth
experiments. The DGC initially contained 45 mL of deionized water. A stir bar in the
DGC arena kept the water well mixed. The mass balance on sucrose in the DGC arena (S)
'is

dS 34
V E = 1‘8 A(SSource — S) ( )

arena

117

where me is the volume of the arena, A is the area of the membrane (3.06 cmz) available
for mass transfer, and SSW,” is the concentration of sucrose in the source ﬂask. This

equation can be solved with the initial condition S= 0 to yield the following.

In SM =kSAt (35)
Sbulk—S V

arena

Sucrose concentrations in the arena were measured over time by HPLC analysis. The

value of ks was found to be 0.26 cm/hr by ﬁtting the model to data. The linearized model

ﬁt (not shown) had a coefﬁcient of determination of 0.9988.

7.4 Results
7.4.1 Oxygen Limitations in the Immobilization Layer

Because the soybean cells are immobilized in agar, diffusion of oxygen from the
head space could potentially limit the rate of growth. This hypothesis was tested in a
preliminary DGC experiment where plant cells were unifomrly distributed throughout a
single slab of agar. After two weeks, the soybean had grown only near the surface of the
gel in a layer approximately 3 mm thick. This system was closed, providing no
replenishment of oxygen either by exchange of gas head space or reservoir media Based
on these results, plant cells were added only to the 4 mm thick top layer of agar (Layer 3
in Figure 37) in subsequent experiments.

During experiments conducted with a closed head space, the oxygen concentration
in the head space, as measured by gas chromatography, was depleted by over 50% within
1 week (results not shown). To avoid such depletion in subsequent experiments, the head

space was vented to the atmosphere through 0.22 um sterile ﬁlters. Venting allowed the

118

DGC to maintain ambient oxygen concentrations throughout experiments (results not
shown).

7.4.2 Simulation of Unsteady-State Diffusion Gradients.

For the three slab-pouring method, the model predicted 95% of steady-state
would be reached within 2.1 days. The criteria for 95% of steady-state was that every
ﬁnite-element node must be at greater than or equal to 95% of its steady-state value. For
the single-slab method, with zero sucrose concentration throughout the gel, the
corresponding time was 33.1 days. These numbers were the same source concentrations
of both 20 and 30 g/L. Figure 38 shows three time proﬁles for the unsteady-state
solution. The times shown are 1, 5, and 40 hour(s) (top to bottom) after initiation of
diffusion. The side-view of chamber is depicted. The small graph on the top left is a

legend for the sucrose concentration.

'2' ..'
7c. . 1 f 20
. ,' ' .. L
'.‘ . A
-.\ .. -
.

 

Sucrose (g/L)

2.2
1.11 Depth (cm)

 

 

Position (cm)

Figure 38. Simulation of approach to steady-state.

119

7.4.3 Grayscale and Concentration Correlation

Figure 39 shows grayscale values as a function of soybean concentration for 4
brightness settings. Each line and data set corresponds to a different brightness setting.
Concentrations are given in relative units as a percent of soybean suspension (1 rmn
sieve) per total volume of suspension and water. In all cases, ranges existed where the
grayscale values of acquired images were found to increase linearly with cell

concentration. Outside of the

 

250 1-

 
   

\
-rf‘

200 ~-

150 -~

Grayscale

 

Increasing Brightness

 

0% 25% 50% 75%
Concentration

100%

Figure 39. Correlation of image grayscale with plant-cell density.

linear range, the grayscale response becomes black and white saturated at low and high
concentrations, respectively. At the highest brightness setting shown in Figure 39,
grayscale is a linear ﬁmction of concentration ﬁ'om 0 to 40%. Past 40% the response is
saturated. The next three lower settings have linear ranges ﬁ'om 0 to 80%, 20 to 100%,
and 40 to 100% respectively. The four culture types of varied morphologies produced

results similar to those shown in Figure 39 although the slopes varied.

120

7.4.4 Growth Studies
7.4.4.1 Linearity of Diffusion Gradients

Example total sugar assays from soybean growth experiments are shown in Figure
40 for a 0 to 20 g/L gradient and in Figure 41 for a 0 to 30 g/L gradient. Results are shown
for plugs taken on both sides of the centerline. The gradients predicted by the model are
also shown in Figure 40 and Figure 41. At the source end of the gradient, the model agreed
well with experimental data. However, total sugar assays were higher than model
predictions toward the sink end of the gradient. The cause of this trend was not
determined although interference by medium components was experimentally disproved.

The total sugar assay of the soybean media without sucrose produced no absorbance.

20

 

.a
0'!

    

 

Total Sugar (gIL)
8

0'1

1: Source Outlet Side,
0 Source Inlet Side i
Model

 

 

 

 

 

 

0 1 2 3 4 5
Position (cm)

Figure 40. Total sugar gradient data and model for a 0 to 20 g/L experiment.

Sucrose concentrations in the source and sink ﬂasks were measured for several
experiments in order to verify pseudo-steady-state conditions. Losses from source ﬂasks

ranged from 0.08 to 0.41 g/L-week. For six different DGC experiments, source ﬂask

121

losses were 0.26:1:0.l3 g/L-week at a 95% conﬁdence level. A 20 g/L source ﬂask would

therefore have an average decrease in concentration of only 1.3% per week.

 

    

 

 

 

 

 

 

 

30
25
320 —.
a.
a 15 «»
tn
3 1o -- .
'2 D Source Outlet Sidel
5 - o Source Inlet Side 1
Model 1
0 1 1 1 1
0 1 2 3 4 5

Position (cm)

Figure 41. Total sugar gradient data and model for a O to 30 g/L experiment.
7.4.4.2 Effect of Sucrose

The effect of sugar concentration on soybean grth was studied in DGC
experiments with reservoir ﬂask concentrations of 0 to 20 g/L and 0 to 30 g/L. Figure 42
shows the ﬁnal day image of the DGC for a O to 20 g/L experiment. Figure 43 shows the
plot of grayscale values versus position on various days for the same experiment. As in
all results shown, the source reservoir is on the right (y=5 cm), and the sink reservoir is
on the left (y= 0 cm). The dark to light trend from left to right depicts increasing soybean
concentration and grayscale. The series of grayscale plots shows both the spatial and
temporal dependence of plant cell concentration. A maximum in concentration and thus

growth rate is seen near the source reservoir.

122

 

Figure 42. Image from day 13 of a 0 to 20 g/L sucrose experiment.

 

 

 

 

120

100 ~- ' 13 days
2 80 __ 11 days
3 60 -. 8 days
2
0 40 -- MM“ 1day

20 -'
0

0 1 e .4. 4 5
Position (cm)
Figure 43. Grayscale proﬁles for a 0 to 20 g/L experiment.

Decreased soybean grth at higher sucrose concentrations was observed in one
of the O to 20 g/L experiments and all of the 0 to 30 g/L experiments. Figure 44 shows an
example grayscale obtained for a 0 to 30 g/L experiment. The maximum cell concentration
occurred slightly past 1 cm from the sink reservoir. Based on total sugar assays of the gel,

a concentration of approximately 14 g/L was determined to be optimal for growth

123

(maximum cell concentration). Essentially no growth occurred at concentrations above 20
g/L.

50

 

40 -~

30 ~-

20 «-

Grayscale

10-~

 

 

 

0 1 2 3 4 5
Position (cm)

Figure 44. Grayscale proﬁle from day 14 of a 0 to 30 g/L experiment.

Figure 45 and Figure 46 show the chamber image and ﬁnal grayscale plots for a
second 0 to 30 g/L experiment. In contrast to other experiments, the plant cell
concentration also varied perpendicular to the gradient. Plant cell concentrations were
higher in the outlet half of the chamber at high sucrose concentrations, as seen in both the
chamber image and grayscale plots of each side. This result is explained by total sugar
assay data. Fight 47 shows total sugar concentrations for the experiment. While the
gradient was linear on the source outlet side, the gradient did not appear to be linear on
the source inlet side. Furthermore, past roughly the 3 cm position the total sugar
concentrations were signiﬁcantly higher on the source inlet side. The outlet side of the
chamber in which total sugar concentrations reached less than 20 g/L had growth all the

way to the source-reservoir edge of the DGC. The inlet side, however, reached over 25

124

g/L, and growth stopped at approximately 20 g/L. Maximum growth appeared to have

occurred at approximately 10 g/L total sugar.

 

Figure 45. Image from day 14 of a 0 to 30 g/L sucrose experiment.

 

120
100 -»
80 ~-
60 --
40 -~

 

 

Grayscale

 

Outlet Side
Inlet Side

20 _.

 

 

 

 

 

 

 

0 1 2 3 4 5
Position (cm)

Figure 46. Grayscale proﬁles for inlet and outlet halves of the DGC.

125

 

0)
O

 

DOutlet Side 1
o Inlet Side

N
0|
0

 

 

 

 

 

N
915» 0
(I) :1
E10» 0 a
.2 9 o

5-1

0 1 1 .

0 1 2 3 4 5

Position (cm)

Figure 47. Total sugar assays for inlet and outlet halves of the DGC.

7.4.4.3 Effect of l-BS Medium Components

Three components of Gamborg 1-B5 medium were tested for their effect on
soybean growth: basal salt mixture, 2,4-D, and ammonium sulfate. Figure 48 shows the
ﬁnal day image and Figure 49 shows the grayscale plot for a basal salt experiment with
sink and source ﬂask concentrations of 0 g/L and 9.3 g/L respectively. Soybean growth
was signiﬁcantly affected by salt concentration. Maximum growth occurred around the
recommended concentration (3.1 g/L) used in 1-B5 medium. Less marked results were
seen with the 2,4-D experiment ranging from 0 to 2 ppm (results not shown). Maximum
growth occurred near the center of the chamber which corresponds to the normal
concentration of 2,4-D (1 ppm) in the Gamborg 1-B5 medium. The experiment utilizing
an ammonium sulfate gradient of 0 to 0.67 g/L showed little effect of nitrogen
concentration on growth (results not shown). The gradient ranged from O to 5 times the

normal medium concentration in the sink and source ﬂasks respectively. The grayscale

126

plot on the ﬁnal day was essentially ﬂat except for somewhat lower values near the sink
reservoir. These results indicate that nitrogen levels above that in the normal 1-B5

medium did not enhance or retard growth.

 

Figure 48. Image from day 13 of Gamborg basal salts experiment.

70
60 ~-
50 -»
40 -»
30 ~-
20 --
10 --
0 1 4 1 p
0 1 2 3 4 5
Position (cm)

 

Grayscale

 

 

 

Figure 49. Grayscale proﬁle from day 13 of Gamborg basal salts experiment.

127

7.4.4.4 Effect of Antibiotics

Figure 50 shows the results for a Timentin gradient ranging from O to 500 ppm in
the sink and source ﬂasks respectively. Soybean concentration appears to diminish
beyond the 2 cm position. Figure 51 shows grayscale results for replicate experiments
using a 0 to 300 ppm ampicillin gradient. While the magnitude of the grayscale varies
between replicates due to camera effects and/or differences in cell density, the grayscale
trends with respect to position are in good agreement. Ampicillin appears to inhibit
growth around the 4 cm position. Figure 52 shows a comparable experiment for a 0 to 600
ppm ampicillin gradient. In this experiment, inhibition appears to start at the center of the

chamber.

 

N
O

Grayscale

1—8—8

0000

l
r

 

 

 

2 3 4 5
Position (cm)

0
.3

Figure 50. Grayscale proﬁle for day 11 of a Timentin experiment.

128

 

140
120 ~1
100 -~
80 0
60 -~
40 --
20 -~

0 1 1 1 1
0 1 2 3 4 5
Position (cm)

Grayscale

 

 

 

 

Figure 51. Grayscale proﬁles for day 15 of replicate 300 ppm ampicillin experiments.

Ampicillin gradients were measured for the 300 and 600 ppm experiments. All
three experiments had essentially linear gradients of ampicillin (results not shown).
However, these concentrations were well below expected values. The average
concentration for the 300 ppm experiments at the 4.5 cm position was 84 ppm. The
average concentration for the 600 ppm experiment at the 4.5 cm position was only 34
ppm. Values much closer to source ﬂask concentrations were expected. A signiﬁcant
difference between these experiments was their duration. The 300 ppm experiments
lasted 15 days while the 600 ppm experiment lasted 23 days. Assays of the reservoir
source ﬂasks indicated an average decrease of 43% for the 300 ppm experiments and a

decrease of 54% for the 600 ppm experiment.

129

 

Grayscale
00
O

 

 

0 1 1 1 1
0 1 2 3 4 5

Position (cm)

 

Figure 52. Grayscale proﬁle for day 23 of 600 ppm ampicillin experiment.
7.4.5 Isoﬂavonoid Production Studies

Three DGC experiments were run with gradients of ammonium sulfate, which is
the sole source of reduced nitrogen supplied in the Gamborg 1-B5 medium. Daidzein and
genistein concentrations for experiments with source ﬂask concentrations of 1, 3, and 5
times the normal value are shown in Figure 53, Figure 54, and Figure 55, respectively.

Concentrations are reported as ug isoﬂavonoid per g of dry cell mass. Figru'e 53 shows a

clear trend of increasing isoﬂavonoid production with increasing supply of nitrogen. A
similar trend is not seen in Figure 54 or Figure 55. Similarities are limited to a maximum
genistein level at intermediate values. Soybean stock cultures were also different among
the runs however. Soybean suspension culture used in Figure 53 was grown in ammonium

sulfate free medium for the 2 week cycle prior to use.

130

 

 

 

 

 

 

 

 

 

35
A 301.
E . .
g 25 __ DDaIdzeIn
v IGenistein
.5 20»
‘5
.1: 151-
r:
a)
g 10--
o
O 5 «1
1 2 3 4 5
Sink Section Source

Figure 53. Free isoﬂavonoid concentrations for 0 to 1x nitrogen gradient.

 

 

 

 

 

 

 

 

 

 

 

14
A 12.1 DDaldzein
g lGenistein
3- 10.11
s e~
‘65
3: 6-
c
a) —
8 41
o
o 2 _U.
O L I -
1 2 3 4 5
Sink

Section Source

Figure 54. Free isoﬂavonoid concentrations for 0 to 3x nitrogen gradient.

131

 

 

 

 

 

 

 

 

 

 

 

 

 

35
A 30 __ a Daidzein
E IGenistein 1
g 25 «-
Iii
i: 15 ~~
:
a)
g 10 ~-
8
5 4. -—-I—‘— I j—
0 -—
1 2 3 4 5
8m" Section SW09

Figure 55. Total isoﬂavonoid concentrations for 0 to 5x nitrogen gradient.

Two other medium components were tested, thiamine and 2,4-D. The results for
the thiamine experiment are shown in Figure 56. A trend of increasing isoﬂavonoid
concentrations with thiamine is illustrated. Isoﬂavonoids were not detected in sections 1
and 2, explained in part by the low quantity of tissue which was available for extraction.
Figure 57 shows the results of the 2,4-D experiment with a source reservoir concentration
of 5 ppm. Maximum isoﬂavonoid concentrations occur in the second section, where 2,4-
D is expected to be at 1 to 2 ppm.

Phenylalanine, a precursor to daidzein and genistein, was tested in 2 experiments.
Figure 58 and Figure 59 show the results for O to 1.0 g/L and O to 0.1 g/L phenylalanine
experiments, respectively. The O to 1 g/L gradient produced a ﬂat proﬁle for both

isoﬂavonoids. The O to 0.1 g/L gradient produced a maximum however. Production is

132

maximum in the second section, with the greatest contrast between sections 2 and 3. The

maximum is expected to be at approximately 50 ppm phenylalanine.

 

 

 

 

 

 

 

 

 

 

6
E 5 -_ DDaidzein
Q IGenistein
a
v 4 q
r:
.2
'5' 3“
‘3'
o 2 "
c
o
0 1 «
o 4 i i 4.
1 2 3 4 5
Sink Section Source

Figure 56. Free isoﬂavonoid concentrations for O to 1x thiamine gradient.

60

 

 

D Daidzein

5 O ” I Genistein

 

 

 

4o~~ _

30“

20-.

 

Concentration (ppm)

10“

 

 

 

 

 

 

o -_
1 2 3 4 5
Sink Section Source

Figure 57. Total isoﬂavonoid concentrations for 0 to 5x 2,4-D gradient.

133

 

 

 

 

 

30
IDDaidzein
’E‘ 25 -~ :lGenistein
a.
o. ‘—
V 20“ —
c
.9
E 15 *-
“'5
o 10--
c
o
O

 

 

 

 

 

         

Source Middle Sink
Secﬁon

Figure 58. Free isoﬂavonoid concentrations for 0 to 1 g/L phenylalanine gradient.

 

 

 

 

 

 

 

 

7O
60 _. n Daidzein
E lGenistein:
g so —»
25.
E 30 T
8
r: 20 -*
o
O
1 0 ‘-
0 -_
1 2 3 4 5
Sink Section Source

Figure 59. Total isoﬂavonoid concentrations for 0 to 0.1 g/L phenylalanine gradient.

134

7.5 Discussion
7.5.1 Correlation of Grayscale and Concentration

Image acquisition and analysis was found to provide a suitable, non-invasive, on-
line method for monitoring plant cell growth in the DGC. An important feature of the
image acquisition software used in this work was the adjustable brightness setting. As
seen in Figure 39, no single setting allowed for a linear relationship across the entire range
of concentrations. S-curves with black and white saturation at either extreme are seen. B y
adjusting the brightness, however, a linear region can be obtained for the concentration of
interest. A large range of concentrations could potentially be spanned by using multiple
brightness settings. The trade-off for decreased brightness, and therefore the ability to
work in a linear response range at higher concentrations, is a loss of sensitivity. This trend
is seen in Figure 39 as decreasing slopes for decreasing brightness settings. An optimum
setting could be chosen for a particular application by ﬁnding the brightest setting which
gave a linear response over the entire concentration range of interest.

Cell density data from the ﬁnal day of several experiments were collected in order
to supplement grayscale data. Correlation of ﬁnal day cell density and grayscale ﬁom

sections was poor as indicated by low coefﬁcient of determination (1*2 = O.31:l:0.33)

values. While sections of a grayscale image are exactly the same size, DGC gel sectioning
introduced error. Calculations assume that all sections are equal. Residual, unmelted gel
can remain with the plant tissue when dried and weighed, increasing the measured cell
mass. For some experiments, the concentration gradient was shallow across the sections,

yielding a ﬂat plot of grayscale versus concentration, and thus a low coefﬁcient of

135

determination. It should be noted that these problems were not present in the greyscale
and dry weight correlation work described above. Image analysis provided greater
sensitivity and precision within any single experiment than dry weight measurement.
7.5.2 Suitability of DGC for Growth Studies

Sugar gradient data and modeling indicated that the DGC method can approximate
steady-state conditions in order to study plant cell culture growth. Diffusion of sucrose
across the arena and consumption by plant cells, were found to have a minimal impact on
reservoir ﬂask sucrose concentrations. Although non-linear gradients could arise due to
cellular consumption of metabolites, no evidence of such gradients was seen except for
Figure 47 which is likely explained by reservoir non-idealities discussed below.

For the linear gradients shown in Figure 40 and Figure 41, the total sugar
concentrations were higher than model predictions. This result may be explained in part
by extracellular hydrolysis of sucrose to glucose and fructose by cell wall invertases
(Fowler, 1982). The hydrolysis is a rapid process compared to consumption (Hooker and
Lee, 1992) and may, therefore, increase the concentration driving force for sucrose
difﬁrsion across the DGC arena. The total sugar assay used in this study would
collectively quantify the sucrose and its hydrolysis products glucose and fructose.

Measured ampicillin concentrations were signiﬁcantly lower than expected at the
end of the experiments. Degradation and uptake by plant cells are two possible
explanations. In either case, the actual concentrations to which the cells were exposed
were different than those predicted assuming a linear gradient based on initial ampicillin

concentration. The results indicate that whenever possible, assays of the chemoeffector

136

should be conducted when working with a new compound and/or culture to determine the
signiﬁcance of uptake or degradation. In cases where degradation is not signiﬁcant, then
such results would provide information about the rate of cellular uptake.

7.5.3 Non-Ideality of DGC Reservoirs

Several sucrose experiments yielded inlet- and outlet-side concentration gradients
for total sugar that were not equal. Examples of this trend are shown in Figure 40 and
Figure 47. This result is attributed to ﬂow patterns in the reservoir. To prevent swelling
of the gel or ﬂooding of the arena, low reservoir ﬂow rates were used initially. These low
ﬂow rates may have led to concentration gradients in the x-direction within the reservoirs.
In subsequent experiments having a ﬂow rate greater than approximately 6 mL/h, the
problem did not reappear.

7.5.4 Determination of Optimum Growth Conditions

The optimal concentrations of Gamborg salts and 2,4-D were essentially that of
Gamborg l-BS medium. This was not the case for sucrose gradient experiments however,
which suggested that sugar concentrations became inhibitory above 15 g/L to the
immobilized soybean cultures. In the original report on formulation of the 7 B5 medium
(Gamborg et al., 1968), the authors found that soybean cell yields increased with sucrose
concentration up to 20 g/L sucrose. Furthermore, plant cell growth kinetics are commonly
described by a Monod model when sucrose or glucose is the carbon and energy substrate
(Hooker and Lee, 1992; van Gulik et al., 1992; Guardiola et al., 1995; Westgate et al.,
1991; Schnapp et al., 1991). The Monod model is incapable of describing substrate

inhibition. Other reports of sucrose inhibition at this level are unknown to the authors.

137

Cell growth can be studied under steady-state conditions using other experimental
systems such as a CSTR. However, several different steady-states would need to be
achieved to determine the effect of the dependent variable for a range of values. Such
information can be obtained in a single DGC experiment. Genetic instability of some plant
cultures adds complexity to the comparison of multiple experiments. DGC studies yield a
continuum of data from a single inoculum. While multiple replicates could be run in
CSTRs or shake ﬂasks, this approach is expensive due to the demand for equipment or
manpower. Finally, image capture provides a quick and non-invasive method of data
acquisition not available in the other systems.

7.5.5 Determination of Optimum Isoﬂavonoid Conditions

DGC experiments which focused on isoﬂavonoid quantiﬁcation were performed
after the methods for growth studies had been established. The chemoeffectors chosen for
these studies were largely those used in growth studies, with the addition of a vitamin and
an isoﬂavonoid precursor. Nitrogen was chosen for study as its levels affect secondary
metabolism in numerous biological systems. Nitrogen is also present in the amino acid
phenylalanine. The conversion of phenylalanine to cinnamic acid by phenylalanine
ammonium lyase (PAL) is the branch point from primary to secondary metabolism (Barz
and Welle, 1992). As ammonium sulfate is the only reduced nitrogen source, its
availability was expected to be important (Gamborg et al., 1968). By the same logic,
phenylalanine was also tested. It was hypothesized that a small pool of phenylalanine,
and thus a slow conversion by PAL, could be rate limiting in the isoﬂavonoid pathway.

The hormone 2,4-D was chosen as its negative effect on secondary metabolite production

138

has been reported. Thiamine was chosen as the vitamin which is provided in the highest
quantity in the soybean medium.

The nitrogen studies shown in Figure 53, Figure 54, and Figure 55 do not show a
consistent trend such as an optimum level of ammonium sulfate. Three gradient slopes
were tested in search of an optimum level. However, the results indicate that the history
of the cell culture used in the study is important. The experiment illustrated in Figure 53
used cells which had been deprived of ammonium sulfate, while the others did not. The O
to 1x experiment was conducted with the assumption that nitrogen was not rate limiting
under normal conditions, but that a rate limitation could be induced. While the effect was
signiﬁcant for this ﬁrst experiment, additional runs focused on conditions useful for
bioreactor operation. Increased ammonium sulfate levels would only be of value if
isoﬂavonoid production was enhanced for normal (non-deprived) cells. The results in
Figure 54 and Figure 55 indicate that there is not a signiﬁcant beneﬁt from elevated
nitrogen levels.

Supplementation of phenylalanine appeared to have a marginal effect on
isoﬂavonoid production. For the steep 1 g/L gradient shown in Figure 58, the ﬂat proﬁle
is inconclusive. Supplementation may have had no effect. Alternatively, the effect may
have been saturated very close to the sink reservoir. The latter was hypothesized and
tested with the 0.1 g/L experiment. Figure 59 indicates that isoﬂavonoid production is
effected by supplementation. The results also indicate that bioreactor operation may
beneﬁt from supplementation, but not with a large improvement in productivity. While a

comparison of concentrations from these experiments would be useful, this must be done

139

with caution as Figure 58 reports only free isoﬂavonoids while Figure 59 reports total
isoﬂavonoids. In general, comparison between DGC experiments which use different
inoculums must be done cautiously. While replicate growth studies have demonstrated
consistency with DGC methods, variation in plant cell culture can lead to dramatic

differences in isoﬂavonoid levels over time.

7.6 Conclusions

The Diffusion Gradient Chamber (DGC) has been demonstrated as a tool for
studying the effects of chemical concentrations on plant cell culture growth and product
formation. Image acquisition provided a sensitive, non-invasive method for tracking
growth over the course of an experiment, based on the linear relationship between image
grayscale and cell concentration. Medium components and antibiotics were found to
affect soybean cell culture growth. Optimal concentrations for Gamborg B5 salts and 2,4-
D were found to be in the range of normal medium component values. Inhibition of
growth was seen for the antibiotics ampicillin and Timentin, and for total sugar
concentrations above 15-20 g/L. Nitrogen source, phenylalanine, 2,4-D, and thiamine were
found to affect production of daidzein and genistein. Optimal conditions for productivity
appeared to be within recommended medium values excepting a potential beneﬁt from

phenylalanine supplementation.

8. SUMMARY AND CONCLUSIONS

Continuous isoﬂavonoid production from soybean with biocatalyst throughput
was demonstrated in a MFB bioreactor for over 2 months. Plug-ﬂow biocatalyst
movement was observed, utilizing a magnetic valve for solids. Both biocatalyst
microscopy and Thiele modulus calculations indicated that intra-biocatalyst oxygen
diffusion was not signiﬁcant.

Continuous operation has been demonstrated past start-up to the point of steady
growth rates and sugar concentrations. However, modeling indicated that achievement of
steady-state is dependent on the inherently long, liquid residence time. Contamination of
the long-term continuous bioprocess, and plant cell growth on the exterior of the
immobilization matrix remain concerns for process feasibility. A simple control strategy
was able to maintain DO and C02 near target levels. Model results predicted an optimum
integer interval between biocatalyst exchange of 4 days. While isoﬂavonoid productivity
was improved from earlier runs, concentrations were well below current shake ﬂask
levels, due in part to the use of antibiotics. Daidzein, but not genistein, productivity
exceeded the target level indicating commercial potential. Research is needed to better
understand the relationship between isoﬂavonoid production kinetics and bioreactor
process variables.

Based on solenoid power per volume calculations, the MFB bioreactor is feasible
at large scale, but may require the use of a third (gas) phase. If needed, the gas phase could

be provided by microbubbles which do not increase solenoid power consumption like

140

141

traditional sparged bubbles. The MP B scales up favorably with regard to power per
volume requirements. Calculations indicate that an order of magnitude reduction in power
consumption can be achieved over a traditional CSTR fermenter. Added capital costs for
the solenoid were estimated to be reasonable.

The Diffusion Gradient Chamber (DGC) was demonstrated as a tool for studying
the effects of chemical concentrations on plant cell culture growth and product formation.
Image acquisition provided a sensitive, non-invasive method for tracking growth over the
course of an experiment, based on the linear relationship between image grayscale and cell
concentration. The DGC was used for detennining optimal medium constituent
concentrations, and inhibitory levels of toxic compounds. Secondary metabolite
(isoﬂavonoid) accumulation was also studied. Optimal conditions for productivity
appeared to be within recommended medium values excepting a potential beneﬁt from
phenylalanine supplementation. However, DGC studies could be expected to yield
signiﬁcant insight to productivity improvements for a plant bioprocess where medium

optimization studies have yet to be done.

LIST OF REFERENCES

9. LIST OF REFERENCES

Adesanya, S.A., O'Neill, M.J., Roberts, M.F. (1986) Structure-related fungitoxicity of
isoﬂavonoids, Physiol. and Mol. Plant Path. , 29, 95-103.

Aiba, S., Humphrey, A.E., Millis, NP. (1973) Biochemical Engineering, Academic Press,
New York.

Akiyama, T., Ishida, J ., Nakagawa, S., Ogawara, H., Watanabe, S., Itoh, N., Shibaya, M.,
Fukarni, Y. (1987) Genistein, a speciﬁc inhibitor of tyrosine-speciﬁc protein
kinases, J. Biol. Chem, 262, 5592-5595.

Ames, T.T., Thompson, V.S., Worden, RM. (1995) A Three Phase Magnetoﬂuidized
Bed Bioreactor for Production of Plant Secondary Metabolites, Annual Meeting of
American Institute of Chemical Engineers, Miami Beach, FL, November 14, 1995.

Ames, T.T., Worden, RM. (1997) Continuous Production of Daidzein and Genistein
from Soybean in a Magnetoﬂuidized Bed Bioreactor. Biotech. Prog., in press.

Ananta, M., Subroto, A., Doran, P. (1995) Oxygen transfer and culture characteristics of
self-immobilized Solanum aviculare aggregates, Biotech Bioeng, 47, 541-549.

Andrews, G. (1988) Fluidized-bed bioreactors. Biotechnology and genetic engineering
reviews, 6, 151-169.

Archarnbault, J. (1991) Large-scale (20 L) culture of surface-immobilized Catharanthus
roseus cells, Enzyme Microb. Technol, 13, 882-892.

Atkinson, B., Mavituna, F. (1983). Biochemical Engineering and Biotechnology
Handbook, Stockton Press, New York, 263.

Bailey, J .E., Ollis, DP. (1986) Biochemical Engineering Fundamentals, McGraw Hill,
New York.

Balaban, R.S., Mandel, L.J. (1990) Optical methods for the study of metabolism in intact
cells, Noninvasive Techniques in Cell Biology, Eds. J .K. Foskett and S. Grinstein,
Wiley-Liss, New York, 213-236.

Barz, W., Welle, R. (1992) Biosynthesis and metabolism of isoﬂavones and pterocarpan
phytoalexins in chickpea, soybean and phytopathogenic fungi, Phenolic
Metabolism in Plants, Ed. H.A. Stafford and R.K. Ibrahim, Plenum Press, New
York, 139-165.

Bate, NJ. (1994) Quantitative relationship between phenyalanine ammonia-lyase levels
and phenylpropanoid accumulation in transgenic tobacco identiﬁes a rate-

142

143

determining step in natural product synthesis, Proc. Natl. Acad. Sci., 91, 7608-
7612.

Bramble, J.; Graves, D.; Brodelius, P. Plant cell culture using a novel bioreactor: the
magnetically stabilized ﬂuidized bed. Biotechnology Progress 1990, 6, 452-457.

Bramble, J .L., Graves, D.J., Brodelius, P. (1990) Plant cell culture using a novel
bioreactor: The magnetically stabilized ﬂuidized bed, Biotechnol. Prog. , 6, 452-
457.

Bredwell, M.D., Telgenhoff, M.D., Worden, RM. (1994) Enhancement of gas mass
transfer using microbubble dispersions, Annual Meeting of American Institute of
Chemical Engineers, San Fransico, CA, November 13-18, 1994.

Bredwell, M.D., Telgenhoff, M.D., Worden, R.M. (1995a) Formation and coalescence
properties of microbubbles, App]. Biochem. Biotechnol. ,51/52, 501-509.

Bredwell, M.D., Telgenhoff, M.D., Worden, R.M. (1995b) Effects of Using Microbubble
Dispersions on Synthesis Gas Fermentations, Annual Meeting of American
Institute of Chemical Engineers, Miami Beach, FL, November 14, 1995.

Brodelius, RE. (1990) Transport and accumulation of secondary metabolites, Progress in
Plant Cellular and Molecular Biology, Proceedings of the VIIth International
Congress on Plant Tissue and Cell Culture, Eds. H.J.J. Nijkamp, L.H.W. van der
Plas and J. van Aartrijk, Kluwer Academic Publishers, Boston.

Brown, D., Groom, C., Vitanik, M., Brown, M., Cooper, J ., Arditti, J. (1982) Effects of
fungicides and bactericides on orchid seed germination and shoot tip cultures in
vitro, Plant Cell Tissue Organ Culture, 1, 165-180.

Caboche, M., Lark, KB. (1981) Preferential replication of repeated DNA sequences in
nuclei isolated from soybean cells grown in suspension culture, P.N.A.S., 78,
1 73 1 -l 73 5.

Chappell, J ., Halbrock, K. (1984) Transcription of plant defense genes in response to UV
light or fungal elicitor, Nature, 311, 76-78.

Cho, G.H., Kim, D., Pedersen, H., Chin, C. (1988) Ethephon enhancement of secondary
metabolite synthesis in plant cell cultures, Biotechnol. Prog., 4, 184-188.

Cho, M.J., Harper, J .E. (1991) Effect of inoculation and nitrogen on isoﬂavonoid
concentration in wild-type and modulation-mutant soybean roots, Plant Physio]. ,
95, 43 5-442.

Christou, P. (1994) The biotechnology of crop legumes. Euphytica, 74, 165-185.

144

Cussler, EL. (1984) Drﬂusion: Mass transfer in ﬂuid systems. Cambridge University
Press, Cambridge.

Darvill, A.G., Albersheirn, P. (1984) Phytoalexins and their elicitors: A defense against
microbial infection in plants, Ann. Rev. Plant Physio]. , 35, 243-275.

Dorgelo, E., van der Meer, A., Wesselingh, J. (1985) Measurement of the axial dispersion
of particles in a liquid ﬂuidized bed applying a random walk method, Chem. Eng.
Sci, 40, 2105-2111.

Domenburg, H., Knorr, D. (1995) Strategies for the improvement of secondary metabolite
production in plant cell cultures, Enzyme and Microbial Technology, 17, 674-684.

Dougall, D.K. (1986) Accumulation of natural products in plant tissue cultures revisited,
New Zealand J. T echnol., 2, 71-76.

Dubois, M., Gilles, K., Hamilton, J.K., Rebers, RA, Smith, F. (1956) Anal. Biochem.,
28,350.

Einsele, A. (1976) Scaling of bioreactors, theory and reality, 5th Int. Ferm. Symp., Berlin,
Paper 4.13.

Elridge, A.C., Kwolek, W. (1983) Soybean isoﬂavones: Effect of environment and variety
on composition, J. Agric. Food Chem, 31, 394-396.

Emerson, D., Worden, R.M., Breznak, J .A. (1994) A diffusion gradient chamber for
studying microbial behavior and separating microorganisms, Appl. Environ.
Microbiol, 60, 1269-1278.

F acchini, P., DiCosmo, F. (1991) Plant cell bioreactor for the production of
photoberberine alkaloids from immobilizaed Thalictrum rugosum cultures,
Biotech Bioeng., 37, 397-403.

Fett, W.F., Zacharius, RM. (1982) Bacterially induced glyceollin production in soybean
cell suspension cultures, Plant Sci. Lett, 24, 303-309.

Fogler, HS. (1986) Elements of Chemical Reaction Engineering, Prentice-Hall, New
Jersey.

Fogler, HS. (1992) Elements of chemical reaction engineering, 2nd edition, Prentice Hall,
Englewood Cliffs, New Jersey.

Fotsis, T., M. Pepper, H. Adlercreutz, G. Fleischmann, J. Hase, R. Montesana, and L.
Schweigerer (1993) Genistein, a dietary-derived inhibitor of in vitro angiogenesis,
Proc. Nat]. Acad Sci, 90, 2690-2694.

145

Fotsis, T., Pepper, M., Adlercreutz, H., Fleischmann, G., Hase, J ., Montesana, R.,
Schweigerer, L. (1993) Genistein, a dietary-derived inhibitor of in vitro
angiogenesis. Proc. Nat]. Acad. Sci. , 90, 2690-2694.

Fowler, M.W. (1982).]. Chem. T ech. Biotechnol., 32, 338-346.

Freund, V.S., Worden, RM. (1990) Gas hold-up in a magnetically stabilized three-phase
ﬂuidized bed bioreactor, Annual Meeting of American Institute of Chemical
Engineers, Chicago, IL, Nov 11-16.

Froment, G., Bischoff, K. (1979) Chemical Reactor Analysis and Design, John Wiley &
Sons: New York.

Gamborg, 0., Miller, R., Ojirna, K. (1968) Nutrient requirements of suspension cultures
of soybean root cells, Experimental Cell Research, 50, 151-158.

Gazaway, W.S., Hagan, AK. (1989) Sclerotium blight, In: Compendium of Soybean
Diseases, 3rd ed., Eds. J .B. Sinclair, and RA. Backman, APS Press, St. Paul,
Minn, 48-49.

Graham, M.Y., Graham, TL. (1991) Rapid accumulation of anionic peroxidases and
phenolic polymers in soybean cotyledon tissues following treatment with

Phytophthora megasperma f. sp. Glycinea wall glucan, Plant Physio], 97, 1445-
1455.

Graham, T., Kim, J ., Graham, M. (1990) Role of constitutive isoﬂavone conjugates in the
accumulation of glyceollin in soybean infected with Phytophthora megasperma,
Mo]. Plant-Microbe Interact. , 3, 157-166.

Graham, TL. (1991) A rapid, high resolution high performance liquid chromatography
proﬁling procedure for plant and microbial aromatic secondary metabolites, Plant
Physio], 95, 584-593.

Graham, T.L. (1991b) Flavonoid and isoﬂavonoid distribution in developing soybean
seedling tissues and in seed and root exudates, Plant Physio], 95, 594-603.

Guardiola, J ., Iborra, J .L., Canovas, M. (1995) A model that links grth and secondary
metabolite production in plant cell suspension cultures, Biotechnol. Bioeng, 46
291-297.

Hahlbrock, K., Lamb, C.J., Purwin, C., Ebel, J., Fautz, E., Schafer, E. (1981) Rapid
response of suspension-cultured parsley cells to the elicitors from Phytophthora
megasperma var. sojae, Plant Physio], 67, 768-773.

146

Holden, M.A., Yeoman, M.M. (1987) Optimization of product yield in immobilised
plant cell cultures, Bioreactors and Biotransformations, Eds. G.W. Moody and
RB. Baker, Elsevier Applied Science Publishers, New York.

Homberg, H., Geiger, H. (1980) Fluorescence and structure of ﬂavones, Phytochemistry,
19, 2443-2449.

Hooker, B.S., Lee, J .M. (1992) Application of a new structured model to tobacco cell
cultures, Biotech. Bioeng, 39, 765-774.

Hu, T., Wu, J. (1987) Study on the characteristics of a biological ﬂuidized bed in a
magnetic ﬁeld, Chem. Eng. Res. Des., 65, 23 8-242.

Hubbard, D.W. (1987) Scale-up strategies for bioreactors, Biotechnology Processes:
Scale-up and Mixing, Ho, C.S., Oldshue, J .Y. eds., American Institute of Chemical
Engineers, New York.

Jacobsen, B.J., Meronuck, RA. (1989) Postharvest Pathology, In: Compendium of
Soybean Diseases, 3rd ed., Eds. J .B. Sinclair, Backman, P.A., APS Press, St. Paul,
Minn., 80-82.

Jaraiz-M., E., Wang, Y., Zhang, G., Levenspiel, O. (1984) Theory and operational
characteristics of the magnetic valve for solids, AIChE J. , 30, 951-966.

Jaraiz-M., B, Zhang, G., Wang, Y., Levenspiel, O. (1984b) The magnetic distributor-
downcomer (MDD) for ﬂuidized beds, Power Techno], 38, 53-61.

J ende-Strid, B. (1993) Genetic control of ﬂavonoid biosynthesis in barley, Hereditas,
119, 187-204.

J ing, Y., Nakaya, K., Han, R. (1993) Differentiation of promyelocytic leukemia cells HT.-
60 induced by daidzein in vitro and in vivo, Anticancer Res., 13, 1049-1054.

J ing, Y., Waxman, S. (1995) Structural requirements for differentiation-induction and

growth-inhibition of mouse erythroleukemia cells by isoﬂavones, Anticancer Res. ,
15, 1147-1152.

Kalk, J .P., Langlykke, A.F. (1986) Cost estimation for biotechnology projects, Manual of
Industrial Microbiology & Biotechnology, Demain, A.L., Solomon, N.A. eds.,
American Society for Microbiology, Washington DC.

Kaster, J .A., Michelsen, D.L., Velander, W.H. (1990) Increased oxygen transfer in a yeast
fermentation using a microbubble dispersion, Appl. Biochem. Biotechnol. , 24, 469-
784.

147

Kim, D., Pedersen, H., Chin, C. (1991) Cultivation of Thalictrum rugosum cell
suspension in an improved airlift bioreactor: Stimulatory effect of carbon dioxide
and ethylene on alkaloid production, Biotechnol. Bioeng, 38, 331-339.

Kochs, G., Grisebach, H. (1987) Induction and characterization of a NADPH-dependent
ﬂavone synthase from cell cultures of soybean, Z. Naturforsch, 42, 343-348.

Kosslak, R.M., Bookland, R., Barkei, J ., Paaren, H.E., Appelbaum, ER. (1987) Induction
of Bradyrhizobium japonicum common nod genes by isoﬂavones isolated from
Glycine max, Proc. Nat]. Acad Sci, 84, 7428-7432.

Kudou, S., F leury, Y., Welti, D., Magnolato, D., Uchida, T., Kitamura, K., Okubo, K.
(1991) Malonyl isoﬂavone glycosides in soybean seeds (Glycine max Merrill),
Agric. Biol. Chem, 55, 2227-2233.

Leckie, F., Scragg, A., Cliffe, K. (1991) Effect of bioreactor design and agitator speed on
the growth and alkaloid accumulation by cultures of Catharanthus roseus, Enzyme
Microb. Techno], 13, 296-305.

Lee, C.W., Shuler, ML. (1991) Different shake ﬂask closures alter gas phase composition
and ajmalicine production in Catharanthus roseus cell suspensions, Biotechnol.
Tech, 5, 173-178.

Lee, H.P., Gourley, L., Duffy, S.W., Esteve, J., Lee, J., Day, NE. (1991) Dietary effects
on breast-cancer risk in Singapore, The Lancet, 337, 1197-1200.

Liao, J .C., Delgado, J. (1993) Advances in Metabolic Control Analysis, Biotechnol. Prog,
9, 221-223.

Lindsey, K., Yeoman, M.M. (1984) The synthetic potential of immobilised cells of
Capsicum ﬁutescens Mill cv. Annuum, Planta, 162, 495-501.

Liu, Y., Hamby, R., Colberg, R. (1991) Fundamental and practical developments of
magnetoﬂuidized beds: a review, Power Technology, 64, 3-41.

Longe, T.A. (1989) Colloidal gas aphrons: Generation, ﬂow characterization and
application in soil and groundwater decontamination, Ph.D. Dissertation, Virginia
Polytechnic Institute and State University.

Maurel, B., Pareilleux, A. (1985) Effect of carbon dioxide on the growth of cell
suspensions of Catharanthus roseus, Biotechnol. Lett, 7, 313-318.

McCabe, W.L., Smith, J.C., Harriott, P. (1993) Unit Operations of Chemical Engineering,
F iﬁh Ed. McGraw-Hill, New York, 350-357.

148

McKelvey, S.A., Gehrig, J.A., Hollar, K.A., Curtis, W.R. (1993) Grth of plant root
cultures in liquid- and gas-dispersed reactor environments, Biotechnol. Prog. , 9,
3 17-322.

Menczel, L., Nagy, F., Kiss, Z.R., Maliga, P. (1981) Streptomycin resitant and sensitive
somatic hybrids of Nicotiana tabacum and Nicotiana knightiana: correlation of
resistance to N. tabacum plastids, Theor. App]. Genet, 59, 191-195.

Messina, M., Barnes, S. (1991) The role of soy products in reducing risk of cancer, J.
Natl. Cancer Inst, 83, 541-545.

Miller, K.J., Hadley, J .A., Gustine, D.L. (1994) Cyclic -1,6-1,3-glucans of
Bradyrhizobiumjaponicum USDA 110 elicit isoﬂavonoid production in the
soybean (Glycine max) host, Plant Physio], 104, 917-923.

Mol, J., de Lange, P., Oostdam, A., van der Plas, L. (1990) Use of genetic engineering to
improve yields in cell cultures, e.g. (anti)sense DNA technology, In Progress in
plant cellular and molecular biology; Nijkamp, H., Van der Plas, L., Van Aartrijk,
J. Eds., Kluwer Academic Publishers: Boston, 712-716.

Montgomery, DB. (1969) Solenoid Magnet Design, Wiley-Interscience, New York.

Mousavi, Y., Aldercreutz, H. (1993) Genistein is an effective stimulator of sex horrnone-
binding globulin production in heptocarcinoma human liver cancer cells and
suppresses proliferation of these cells in culture, Steroids, 58, 301-304.

Mueller, S., Yeh, Y., Chen, W.-T. (1992) Tryosine phosphorylation of membrane protein
mediates cellular invation by transformed cells, J. Cell Biol. , 119, 1309-1325.

Mustelin, T., Burn, P. (1993) Regulation of src family tyrosine kinases in lymphocytes,
T IBS, June, 215-220.

Naim, M., Gestetner, B., Zilkah, S., Birk, Y., Bondi, A. (1974) Soybean isoﬂavones:
Characterization, determination, and antifungal activity, J. Agr. Food Chem. , 22,
806-809.

Okada, W., Fukuda, H., Morikawa, H. (1981) Kinetic expressions of ethanol production
rate and ethanol consumption rate in baker’s yeast cultivation, J. Ferment.
Techno], 59, 103-109.

Okura, A., Arakawa, H., Oka, H., Yoshinari, T., Monden, Y. (1988) Effect of genistein on
topoisomerase activity and on the growth of [Val 12]Ha-ras-transformed NIH
3T3 cells, Biochem. Biophysical Research, 157, 183-189.

149

Orr, J .D., Edwards, R., Dixon, RA. (1993) Stress responses in alfalfa (Medicago sativa
L.), Plant Physio], 101, 847-856.

Panda, A., Mishra, S., Bisaria, V., Bhojwani, S. (1989) Plant cell reactors-~a perspective,
Enzyme Microb. Techno], 11, 386-397.

Partridge, J .E., Keen, N.T. (1977) Soybean phytoalexins: Rates of synthesis are not
regulated by activation of initial enzymes in ﬂavonoid biosynthesis,
Phytopathology, 67, 50-55.

Perry, R.H., Green, D.W., Maloney, J .O. (1984) Perry ’s chemical engineers’ handbook,
6th edition, McGraw-Hill, New York.

Pollock, K, Barﬁeld, D., Shields, R. (1983) The toxicity of antibiotics to plant cell
cultures, Plant Cell Reports, 2, 36-39.

Porter, P.M., Banwart, W.L., Hassett, J .J . (1986) Phenol acids and ﬂavonoids in soybean
root and leaf extracts, Environ and Expt]. Bot. , 26, 65-73.

Rao, S. (1990) Root ﬂavonoids, The Botanical Review, 56, 2-85.

Reinhard, E., Kreis, W., Barthlen, U., Hehnbold, U. (1989) Semicontinuous cultivation of
Digitalis lanata cells: production of B-methyldigoxin in a 300-L airlift bioreactor,
Biotechnology and Bioengineering, 34, 502-508.

Rosenshine, I., Duronio, V., Finlay, B. (1992) Tyrosine protein kinase inhibitors block
invasin-promoted bacterial uptake by epithelial cells, Infection and Immunity, 60,
22 1 l -221 7.

Rosensweig, RE. (1978) Process for operating a magnetically stabilized bed, US. Patent
4 115 927.

Rosensweig, RE. (1979) Fluidization: hydrodynamic stabilization with a magnetic ﬁeld,
Science, 204, 57-60.

Rosensweig, RE, Siegell, J .H., Lee, W.K., Mikus, T. (1981) Magnetically stabilized
ﬂuidized solids, AIChE Symposium Series, 77, 8.

Sahai, 0., Knuth, M. (1985) Commercializing plant tissue culture processes: Economics,
problems and prospects, Biotechnol. Prog., 1, 1-9.

Schlatmann, J.E., Moreno, P.R., Vinke, J.L., ten Hoopen, H.J., Verpoorte, R., Heijnen,
J .J . (1997) Gaseous metabolites and the ajmalicine production rate in high density
cell cultures of Catharanthus roseus, Enzyme Microb. Techno], 20, 107-115.

150

Schlatrnann, J.E., Nuutila, A.M., van Gulik, W.M., ten Hoopen, H.J.G., Verpoorte, R.,
Heijnen, J .J . (1993) Scaleup of ajmalicine production by plant cell cultures of
Catharanthus roseus, Biotech. Bioeng, 41, 253-262.

Schmidt, S. (1995) Development of Novel Methods to Measure Random and
Chemotactic Microbial Motility at the Community Level, Studienarbeit, Istitut
fur Verfahrenstechnik, Supervising Professor at Michigan State University: R.
Mark Worden, East Lansing MI.

Schnapp, S. R., Curtis, W. R., Bressan, R. A., Hasegawa, P. M. (1991) Estimation of
growth yield and maintenance coefﬁcient of plant cell suspensions, Biotech.
Bioeng., 38, 1131-1136.

Schneckenburger, H., Renter, B.W., Schoberth, SM. (1985) Fluorescence techniques in
biotechnology, Trends in Biotechnology, 3, 257-261.

Scholar, E.M., Toews, ML. (1994) Inhibition of invasion of murine mammary carcinoma
cells by the tyrosine kinase inhibitor genistein, Cancer Lett. , 87, 159-162.

Sebba, F. (1987) Foams and Biliquid Foams - Aphrons, John Wiley and Sons, Ltd., New
York.

Shuler, M. (1981) Production of secondary metabolites from plant tissue culture--
problems and prospects, Ann. N. Y. Acad. Sci, 65-79.

Shuler, M., Hirasuna, T., Prince, C., Bringi, V. (1990) Bioreactor considerations for
producing ﬂavors and pigments ﬁom plant tissue culture, In Biotechnology and
food process engineering, Scwartzberg, H., Rao, M., Eds., Marcel Dekker, New
York, 45-65.

Siegell, J.H. (1988) Magnetically frozen beds, Powder Technology, 55, 127-132.

Smith, J.M., Davison, S.W., Payne, GP. (1990) Development of a strategy to control the
dissolved concentrations of oxygen and carbon dioxide at constant shear in a plant
cell bioreactor, Biotechnol. Bioeng, 35, 1088-1101.

Takano, H., Furu-une, H., Burgess, J.G., Manabe, E., Hirano, M., Okazaki, M.,
Matsunaga, T. (1993) Production of ultraﬁne calcite particles by coccolithophorid

algae grown in a biosolar reactor supplied with sunlight, App. Biochem.
Biotechnol. , 39, l 59- 1 67.

Taya, M., Yoyama, A., Kondo, D., Kobayashi, T., Matsui, C. (1989) Growth
characteristics of plant hairy roots and their cultutres in bioreactors, J. Chem. Eng.
Japan, 22, 84-89.

151

Ten Hoopen, H., Van Gulik, W., Meijer, J. (1990) Possibilities, problems, and pifalls of
large-scale plant cell cultures, In Progress in plant cellular and molecular biology,
Nijkamp, H., Van der Plas, L., Van Aartrijk, J. Eds., Kluwer Academic Publishers,
Boston, 673-681.

Thompson, V., Worden, R. (1997) Phase holdup, liquid dispersion, and gas-to-liquid
mass transfer measurements in a three-phase magnetoﬂuidized bed, Chem. Eng.
Sci, 52, 279-295.

Thompson, VS. (1993) Characterization of a three-phase magnetically stabilized
ﬂuidized bed bioreactor, Ph.D. Dissertation, Michigan State University, East
Lansing, M.

Thompson, V.S., Worden, RM. (1991) Axial dispersion in a three-phase, magnetically
stabilized bioreactor, Annual Meeting of American Institute of Chemical Engineers,
Los Angeles, CA, November 17, 1991.

Thompson, V.S., Worden, RM. (1992) The effect of tracer diffusion on liquid axial
dispersion in a three phase ﬂuidized bed bioreactor, Chem. Eng. Sci, 47, 3435-
3441.

Uchimiya, H., Handa, T., Brar, D. (1989) Transgenic plants, Journal of biotechnology, 12,
1-20.

Urushibara, S., Kitayama, Y., Watanabe, T., Okuno, T., Watarai, A., Matsumoto, T.
(1992) New ﬂavonol glycosides, major determinants inducing the green
ﬂuorescence in the guard cells of Allium cepa, Tetrahedron Letters, 33, 1213-1216.

van Gulik, W. M., ten Hoopen, H. J ., Heijnen, J. J. (1992) Kinetics and stoichiometry of
grth of plant cell cultures of Catharanthus roseus and Nicotiana tabacum in
batch and continuous fermentors, Biotech. Bioeng. , 40, 863-874.

van Gulik, W.M., Nuutila, A.M., Vinke, K.L., ten Hoopen, H.J., Heijnen, J .J . (1994)
Effects of carbon dioxide, air ﬂow rate, and innoculation density on the batch

growth of Catharanthus roseus cell suspensions in stirred fermentors, Biotechnol.
Prog., 10, 335-339.

van Gulik, W.M., ten Hoopen, H.J., Heijnen, J .J . (1993) A structured model describing
carbon and phosphate limited growth of Catharanthus roseus plant cell
suspensions in batch and chemostat culture, Biotechnol. Bioeng., 41, 771-780.

van't Riet, K. (1983) Mass transfer in fermentation, Trends in Biotechnology, 1, 113-119.

152

Vieth, W. (1994) Gene expression with plant cells, In Bioprocess engineering: kinetics,
mass transport, reactors and gene expression, John Wiley & Sons, New York,
265-3 24.

Voirin, B. (1983) UV spectral differentiation of 5-hydroxy- and 5-hydroxy-3-
methoxyﬂavones with mono-(4';), di-(3',4') or tri-(3',4',5')-substituted b rings,
Phytochemistry, 10, 2107-2145.

Wang, H., Murphy, RA. (1994) Isoﬂavone composition of American and Japanese
soybeans in Iowa: Effects of variety, crop year, and location, J. Agric. Food
Chem, 42, 1674-1677.

Wang, H., Murphy, P.A. (1994b) Isoﬂavone Content in Commercial Soybean Foods, J.
Agric. Food Chem, 42, 1666-1673.

Watanabe, T., Kondo, K., Oishi, M. (1991) Induction on in vitro differentiation of mouse
erythroleukemia cells by genistein, an inhibitor of tyrosine protein kinases,
Cancer Research, 51, 764-768.

Waugh, R., Brown, J. (1991) Plant gene structure and expression, Plant genetic
engineering, Grierson, D. Ed; Chapman and Hall, New York, 1-37.

Wei, H., Bowen, R, Cai, Q., Barnes, 8., Wang, Y. (1995) Antioxidant and
antipromotional effects of the soybean isoﬂavone genistein, Proc. Soc. Exp. Biol.
Med, 208, 124-130.

Wei, H., Wei, L., F renkel, K., Bowen, R., Barnes, S. (1993) Inhibition of tumore
promoter-induced hydrogen peroxide formation in vitro and in vivo by genistein,
Nutrition and Cancer, 20, 1-12.

Weng, D., Chen, L., Han, Y., Zhu, W., Xu, 8., Ouyang, F. (1992) Continuous ethanol
fermentation in a three-phase magnetic ﬂuidized bed bioreactor, AIChE
Symposium Series on F luidized Processes, 289, 107-115.

Westgate, P. J ., Curtis, W. R., Emery, A. H., Hasegawa, P. M. and Heinstein, P. F.
(1991) Approximation of continuous growth of Cephalotaxus harringtonia plant
cell cultures using fed-batch operation, Biotech. Bioeng., 38, 241-246.

Westgate, P.J., Emery, A.H., Hasegawa, P.M., Heinstein, PF. (1991) Grth of

Cephalotaxus harringtonia plant-cell cultures, Appl. Microbiol. Biotechnol., 34,
798-803.

Westlin, B., Axelsson, A. (1991) Diffusion in gels containing immobilized cells: a critical
review, Biotech. and Bioeng. , 38, 439-446.

153

Whitney, P.J. (1992) Novel bioreactors for the growth of roots transformed by
Agrobacterium rhizogenes, Enzyme Microb. Tech. , 14, 13-17.

Widman, M., Emerson, D., Chiu, C., Worden, R., Breznak, J ., Hoppensteadt, F. (1995)
Modeling of microbial chemotaxis in a diffusion gradient chamber, Annual Meeting
of American Institute of Chemical Engineers, Miami Beach, FL, November 14,
1995.

Widman, M.T., Emerson, D., Chiou, C., Worden, RM. (1997) Modeling microbial
chemotaxis in a diffusion gradient chamber, Biotech. Bioeng. , in press.

Worden, R.M., Donaldson, TL. (1987) Dynamics of a biological ﬁxed ﬁlm for phenol
degradation in a ﬂuidized bed bioreactor, Biotech. and Bioeng., 30, 398-412.

Worden, R.M., Emerson, D., Breznak, J .A. (1993) Quantitative analysis of chemotaxis in
a Diffusion Gradient Chamber, Annual Meeting of American Institute of Chemical
Engineers, St. Louis, MO, November 7-12.

Yamada, Y. (1984) Selection of cell lines for high yields of secondary metabolites, Cell
Culture and Somatic Cell Genetics of Plants, 1, 629-636.

Yanagihara, K., Ito, A., Toge, T., Numoto, M. (1993) Antiproliferative effects of
isoﬂavones on human cancer cell lines established from the gastrointestinal tract,
Cancer Res., 53, 5815-5821.

Yang, W., Jaraiz-M., E., Levenspiel, 0., Fitzgerald, TJ. (1982) A magnetic control valve
for ﬂowing solids: Exploratory studies, Ind. Eng. Chem. Process Des. Dev., 21,
71 7-721 .

Yun, D., Hashirnoto, T., Yarnada, Y. (1992) Metabolic engineering of medicinal plants:
transgenic Atropa belladonna with an improved alkaloid composition, Proc. Nat].
Acad Sci. USA, 89, 11799-11803.

Zacharius, R.M., Kalan, EB. (1990) Isoﬂavonoid changes in soybean cell suspensions
when challenged with intact bacteria or fungal elicitors, J. Plant Physio], 135, 732-
736.

Zhong, J ., Seki, T., Kinoshita, S., Yoshida, T. (1991) Effect of light irradiation on
anthocyanin production by suspended culture of Perillaﬁutescens, Biotech.
Bioeng., 38, 653-658.

TE

UNIV. LIBRARIES
llllllllllllllllllllll
812203

21155

MICHran srn
lllllllllllllllllllll“
31293