You are on page 1of 7

Bioprocess and Biosystems Engineering 24 (2002) 273±279

DOI 10.1007/s004490100263

Understanding the bioreactor


G. LideÂn

273
Abstract Analysis of bioreactors is central for successful Introduction
design and operation of biotechnical processes. The bio- The bioreactor can, with some justi®cation, be claimed to
reactor should provide optimum conditions, with respect be the heart of any industrial biotechnical process. In the
to temperature, pH and substrate condition, for example, words of Cooney [1]; ``The continued success of biotech-
besides its basic function of containment. The ability to nology depends signi®cantly on the development of bior-
control the substrate concentration is an important func- eactors, which represents the focal point for interaction
tion of the bioreactor. The substrate concentration can be between the life scientist and the process engineer''.
subject to spatial variation ± advertently or inadvertently ± However, while the progress in molecular microbiology
and may also change with time in batch or fed-batch has been very apparent in the past decade ± shown not
operation. The cellular metabolism will depend on local least by the emergence of several ``ome-tagged'' disciplines
concentrations in the reactor, as well as on the physio- ± the bioreactor itself has (seemingly) undergone only
logical status of the cell. In order to understand the bio- moderate changes in the past 40 years. One reason for this
reactor operation, cellular metabolism must be considered could be that the design and operation of bioreactors are
together with the ¯ow pro®le and the mass transfer char- already more or less fully understood.
acteristics of the bioreactor. Some fundamental aspects of An alternative explanation, however, is that the opera-
bioreactor operation for yeast and bacterial cultivations tion of a large scale bioreactor is in fact so complex, that it
are discussed in this short review. has been dif®cult to approach the fundamental problems
at all, a view expressed in [2] for example.
Keywords Bioreactor analysis, Fed-batch control, The archetypical large-scale fermentation process today
Gradients is a fed-batch process with a high ®nal density based on as
cheap a substrate as possible. The fed-batch process was
List of symbols originally developed in the 1910s for baker's yeast pro-
cj concentration of compound j (mol/m3) duction [3]. The reason behind the signi®cant improve-
R product formation rate (mol/h) ment of biomass yield in fed-batch operation compared to
r volumetric product formation rate (mol/m3 h) batch operation is that glucose repression and over¯ow
t time (h) metabolism in the yeast Saccharomyces cerevisiae leading
u input process variables to ethanol formation can be avoided in the former [4].
V reactor volume (m3) Over¯ow metabolism is by no means limited to S. cerevi-
x length coordinate (m) siae, but is a rather common phenomena shown by several
z state vector of environmental variables other microorganisms such as Escherichia coli, an indus-
aj stoichiometric coef®cient for j trially very important host for recombinant protein pro-
G exchange coef®cient (m2/s) duction. In E. coli, over¯ow metabolism results in the
q density (kg/m3) formation of acetate [5]. Over¯ow metabolism is one
rc turbulent Schmidt number, i.e. leff/qDeff important reason for using a fed-batch operation. Another
leff effective turbulent viscosity (kg/m s) reason for fed-batch operation is to minimize inhibition
m velocity (m/s) effects. These can be caused by the main carbon source
itself or other compounds or impurities in the feed
medium. Understanding inhibition effects may be very
crucial for designing a successful operating strategy [6]. A
fed-batch operation may also be called for by limitations
Received and accepted: 19 June 2001 in the mass or heat transfer in the reactor. The overall
Published online: 26 September 2001 oxygen or heat transfer rates may set a limit for the
Ó Springer-Verlag 2001 maximum permissible carbon source feed rate. It is not
uncommon that the maximum feed rate is determined by
G. LideÂn over¯ow metabolism in the beginning of fed-batch oper-
Chemical Engineering II, Lund University, ation, but limited by mass or heat transfer towards the end
P.O. Box 124, 221 00 Lund, Sweden
E-mail: Gunnar.liden@chemeng.lth.se of operation.
Tel.: +46 462220862 In large bioreactors spatial inhomogeneities, e.g. with
Fax: +46 46149156 respect to energy dissipation and concentration of oxygen
Bioprocess and Biosystems Engineering 24 (2002)

and/or principal carbon source, will most likely occur. importance, not be further discussed here (for a recent
These inhomogeneities depend on the ¯ow ®eld, the review see [12]). Another important factor is the transfer
position and mode of substrate addition and the reaction of product gases ± predominantly carbon dioxide ± out
kinetics, i.e. the cellular metabolism. Developed tools in of the reactor, since these may be inhibitory to the
computational ¯uid dynamics (CFD) have improved the microorganism [13].
possibilities to resolve macroscopic ¯ow pattern. However, The importance of mammalian cell cultures is
cellular physiology must be considered together with the increasing. In mammalian cell cultures shear rates are
physical transport phenomena. Early experimental studies highly signi®cant, which in turn lead to problems
concerning gradients merely showed the existence or ab- with control of oxygen tension. Due to the slow growth
sence of gradients in large-scale bioreactors. Subsequently, the risks of contamination are also much greater.
overall effects on product yields were studied, and more Innovative bioreactor designs for these kinds of cultures
274
recently also space-resolved intracellular measurements of have recently been introduced, e.g. the spinning
mRNA-levels, for example, have been made in large-scale basket design. Also, for growth of plant cells, shear
bioreactors [7]. In this short review, the basic tasks of the sensitivity may lead to problems with suf®cient oxygen
bioreactor and some fundamental problems of bioreactor transfer [14] so the proper design of the impellers is
operation are discussed. important [15].
In short, the bioreactor should provide an optimum
Basic tasks of the bioreactor environment for the desired microbial reactions. In line
There exist a great variety of bioreactors, both commercial with the ``ome-clature'' of current biotechnology (i.e.
and natural. Examples of the latter include ponds, calf genome, transcriptome, proteome, metabolome), one could
stomachs and termite guts [1, 8]. Most commercial bio- introduce the term ``envirome'' for the state vector con-
reactors ± and some natural ± fall into one of the following taining all relevant extracellular variables, e.g. concentra-
categories; unstirred vessels, stirred vessels, bubble col- tion of all chemical compounds and all relevant physical
umns, airlift reactors, membrane reactors, ¯uidized beds, variables (temperature, shear rate, local velocity, turbu-
or packed beds. Division into different categories may also lence intensity and so on). The interactions between the
be based on the mode of agitation (internal mechanical or ``envirome'' and the other ``omes'' are schematically shown
external pumps) or the prevailing continuous phase (gas in Fig. 1. One may look upon the bioreactor as an envi-
or liquid) [9]. rome generator. The task of the biochemical engineer is
From a commercial point of view, the task of the bio- thus to control the ``envirome'' such that the process is
reactor is simply to minimize production costs. This in- optimized. This should be accomplished by using only a
cludes achieving a high product yield, a high productivity
and ± increasingly important ± a high reproducibility.
This overall task can be decomposed into a number of
subfunctions in different ways (a good analysis is given in
[10]). A typical summary is given in Table 1. One may
note that the basic tasks are not orthogonal, but rather
highly collinear. Several of the tasks, e.g. temperature
control, suspension and dispersion, are for instance
helped by stirring. With respect to gas±liquid mass
transfer, focus is normally put on oxygen transfer.
Obtaining a high oxygen transfer rate ± while maintaining
sterility ± was in fact one of the critical questions in the
early development of the penicillin process in the 1940s.
One could even argue that in solving this problem,
the entire scienti®c discipline of biochemical
engineering emerged [11]. Many good reviews are
available on this subject and the overall mass transfer
characteristics of the bioreactor will, despite its

Table 1. Basic tasks of the bioreactor

Function Fig. 1. Schematic representation of interactions between the cellular


environment (``the envirome'') and the cellular metabolism. The
Containment (ensurance of sterility) biochemical synthetic route (or the information ¯ow) from the
Introduction of gaseous reactants (e.g. oxygen) genome to the metabolome is shown as unbroken arrows. Possible
Introduction of liquid reactants (e.g. carbon source) interactions in that route (e.g. transcription factors or effectors
Removal of gaseous products (e.g. carbon dioxide) binding to the genome, enzymatic feed-back loops, or poisoning
Control of the physical environment (e.g. temperature, shear rate, effects) are shown by dashed arrows. The metabolome interacts with
pH) the envirome via excretion or uptake of metabolites. However, the
Suspension (e.g cells, particulate matter) metabolome does not de®ne the envirome in the same manner as the
Dispersion (two-phase systems) genome de®nes the transcriptome. The arrow between the metabo-
lome and the envirome is dashed to indicate this difference
G. LideÂn: Understanding the bioreactor

very limited number of input variables, which furthermore operation have indeed been reported, e.g. [16]. The
are subject to severe constraints. interest in implementation of periodic operation has so far
been very limited, however, and here we only consider fed-
Optimizing bioreactor operation batch operation. To be able to make any use of Eqs. (3)
and (4) it is necessary to de®ne the relevant state vector z
Steady-state operation and to quantitatively describe the relation between r and z.
The optimization problem to be solved in a continuous The dimensionality of z is of course large, however, kinetic
process can be expressed by Eq. (1). expressions (valid within a limited range of operation)
 max may sometimes be obtained by considering only one
u ˆ …u† …R…u†† …1† element in the state vector; the concentration of a
where ``limiting'' substrate. The functional dependence of r
275
Z Z Z on z can then be investigated and an empirical or
semi-empirical model can be derived. A typical example
Rˆ r…z…x; u††dV …2†
would be a Monod type kinetic equation, including
v perhaps a substrate or product inhibition term. With
The control vector (u) consists of variables that can be known parameters, the variational problem can then be
controlled, such as temperature set-point, pH set-point, solved and a feed pro®le, e.g. for substrate or inducer
substrate feed rate, and stirrer rate, whereas the state addition, can be calculated. Chae et al. [17] studied two
vector (z) contains all relevant state variables in the different constructs for production of chloramphenico-
reactor. In chemostat operation, the problem is thus to lacetyltransferase (CAT) based on the inducers isopropyl
®nd an operating point, u*, such that the state-vector z b-D-thiogalactopyranoside (IPTG) and arabinose. By
(the ``envirome vector'') within the range de®ned by the fusing the CAT protein with the green ¯uorescent protein
domain of u, is optimal for the process. (GFP), they were able to get on-line data for protein
production, although the lag time of cyclization of the GFP
Dynamic operation protein had to be considered [18, 19]. Open-loop optimal
In fed-batch operation, the problem is more dif®cult, since feed pro®les for the inducer addition were calculated for
the control vector u is now a time-dependent function. The both IPTG and arabinose. The authors also discussed the
optimization problem is therefore transformed into a potential of using the GFP ¯uorescence for feed-back
variational problem. control of inducer addition.
Z Control strategies need not be model based (Table 2).
Sometimes, reliable models are simply not yet available
F…u† ˆ R…u…t††dt …3†
and other methods must then be used. Altenbach-Rehm
t et al. [20] made an empirical optimization of the addition
with pro®le of the inducer IPTG for heterologous protein
production in E. coli. A ¯exible feed rate pro®le described
Z Z Z by a double sigmoid function with a ®ve-parameter model
Rˆ r…z…u…t†; x††dV …4† was used and parameters were optimized using a so-
v
called genetic algorithm. The resultant activity of the
model enzyme (GDP- mannose pyrophosphorylase) was
The problem is now to ®nd the function u*(t) which more than doubled after modi®cation of the inducer
maximizes the functional F(u). In principle, a continuous addition pro®le.
process can be also operated in transient mode by, e.g. Another strategy for fed-batch control is to use a
periodic changes of inlet medium composition or strategy based on acquired physiological knowledge,
temperature. Possible improvements by using periodic which is not necessarily formalized mathematically [21].

Table 2. Optimization strategy of introduction of substrate or inducer in fed-batch fermentations

Principle of optimization Comment Recent examples

Empirical No mechanistic model is available. Altenbach-Rehm et al. [20]


Statistical methods can be applied
to ®nd optimum u
Physiological No mathematical model is Ê kesson et al. [22]; Taherzadeh et al. [24]
A
necessarily available. However,
physiological reasoning suggests that
one or several measurable variables
should be kept constant or below a
treshhold value
Model based A model based on a possible mechanism Chae et al. [17]
is available. Theoretical analysis can give
the optimum u
Bioprocess and Biosystems Engineering 24 (2002)

 
Examples of physiological control are to hold a substrate @  @  @ @cj
concentration at or below a threshhold value, to main- cj ‡ vi cj ˆ C ‡ aj r …5†
@t @xi @xi @xi
tain a speci®c growth rate at or below a critical value, or
to operate at a particular value of the respiratory quo- leff
tient. One signi®cant advantage of this kind of control Cˆ …6†
qrc
strategy is that it is often readily amenable to feed-back
control implementation. An interesting variant of phys- (The summation convention has been used above for re-
iological control is the so-called probing control [22, 23]. peated indicies.) Since the ¯ow is turbulent in most bio-
In probing control, the physiological status is periodi- reactors, the effective turbulent diffusivity must be
cally probed by making perturbations in the substrate obtained from a model, e.g. the Kolmogorov k± model.
¯ow. By monitoring the response in dissolved oxygen The turbulence model in itself can be questioned, but
276
tension, for instance, it is possible to determine if equally dif®cult ± for reasons discussed previously ± is to
over¯ow metabolism occurs and hence to lower the feed ®nd an expression for the sink term (i.e. the reaction
rate. A similar control strategy was used in [24] for feed- term). In practice, one has to resort to an expression based
rate control of inhibiting hydrolyzates in fed-batch on steady-state kinetics. Calculated concentration pro®les
fermentation using S. cerevisiae. Some of the toxic furan therefore may or may not be fully accurate. Another very
compounds were shown to be converted by yeast cells important complication is the fact that most bioreactors
[25]. The rate of addition of hydrolyzate was controlled are aerated and therefore a turbulent two-phase ¯ow needs
by measuring the response in the carbon dioxide to be modelled. The success in terms of quantitative
evolution rate after step changes in feed-rate. agreement between experimental data (e.g. from laser
Doppler anemometry) and simulation results from such
The problem of gradients modelling has so far been limited. CFD should therefore
In large-scale aerated reactors there will almost certainly still be regarded mainly as a qualitative, but very helpful
be gradients present, both with respect to oxygen and, in tool, which can give clues to the probability of severe
the case of fed-batch or continuous operation, also with concentration gradients.
The in¯uence on process performance has to be ex-
respect to the limiting substrate. This means that the re- perimentally addressed. Although measurements at large
action rate must be known locally (cf. Eqs. 2, 4). What is scales have been made, most experimental studies have
even more problematic is that r will not only depend on been performed using scaled-down systems. An overview
the local value of z, but also on the history of the cell (as of some selected studies is shown in Table 3. Primarily, the
discussed in [26]). This is due to the fact that changes in effects of oxygen or glucose gradients have been investi-
extracellular conditions may trigger regulatory phenome- gated, since these commonly occur. The two most often
na, e.g. gene expression turn-on or turn-off. Obviously, studied microbial systems have been Baker's yeast
this makes it exceedingly dif®cult to model the kinetics of (S. cerevisiae) and E. coli. Both these organisms are of large
product formation during dynamic conditions and the industrial importance and, furthermore, both exhibit
problem must be approached in a different way. One over¯ow metabolism and anaerobic metabolism. They are
suggestion is to consider the four questions given below. therefore sensitive to both glucose excess and to oxygen
(1) What does the ¯ow pattern look like in the reactor, limitation (in fact, not very attractive features for organ-
and is this likely to lead to large gradients? isms used in large-scale production).
Simulations of gradients in a scale-down system have
(2) Will the gradients in¯uence process performance in a been done by step-change experiments in small-scale
negative way? stirred tank reactors (STR), or by connecting two small
(3) What is the probable mechanism behind negative reactors. Step-change experiments in one reactor can be
effects? used to study dynamic effects of a sudden increase in
(4) How should the negative effects be minimized? glucose concentration or aerobic/anaerobic transitions
[38, 39]. A very rapid mixing of glucose can be obtained
The straightforward method of determining presence of in a small STR. However, the oxygen transfer rate is not
gradients is of course to measure concentrations locally in suf®ciently high to enable studies of fast dynamics related
a large-scale reactor [27, 28, 29]. However, for practical to aerobic/anaerobic transitions. A combination of a STR
reasons only a few discrete points can be examined, and and a plug ¯ow reactor may in this case be a better
the picture obtained will therefore be rather incomplete. A option [35]. The a priori expectation is that gradients are
different option is to use mathematical modelling of the likely to have a negative effect on process performance,
¯ow pattern. Computational calculations can be made at which is also reported for biomass yield (cf. Table 3).
different levels of sophistication, from a low degree of However, not all results show decreased performance in
discretisation using compartment models up to fully the large-scale process. The leavening capacity of Baker's
¯edged computational ¯uid dynamic calculations (CFD) of yeast was found to be higher in both a scaled-down
the ¯ow ®eld [26]. For a one-phase ¯ow, the concentration model system and a large-scale process compared to an
®eld for a substance, j, can in principle be calculated from ideally mixed system [36]. Also the stability of a heter-
Eq. (5) in combination with solving the velocity ®eld from ologous protein was found to increase in a non-ideally
the Navier±Stokes equations. mixed system [37].
G. LideÂn: Understanding the bioreactor

Table 3. Overview of some selected investigations concerning effects of substrate gradients in bioreactors

System studied Experimental approach Findings Reference

Effects of oxygen Step-change experiments in one Gluconic acid production Oosterhuis et al. [30]
gradients on gluconic bioreactor (1.6 l); Two connected corresponded to time spent
acid production by lab scale bioreactors (working aerobically
Gluconobacter oxydans volume 1.6 and 0.6 l) ±one
operating aerobically and the
other anaerobically
Effect of oxygen Step change experiments in one Biomass yield decreased and Sweere et al. [31, 32]
gradients in Baker's bioreactor; Continuous cultivation formation of acetic acid
yeast production in two connected bioreactors increased with increasing
(1.5 and 3 l volumes) circulation time
277
Effect of glucose Lab-scale bioreactor (2 l) with No measureable effect on Namdev et al. [33]
gradients in Baker's a recirculation loop. Simulation biomass yield from the glucose
yeast production of log-normal circulation times gradients. Some acetic acid
achieved by controlling substrate formation at long circulation
addition into the recirculation loop times
Effects of oxygen Monte Carlo simulation used Plasmid number was lower in Namdev et al. [34]
gradients on recombinant for switching aeration on/off cells during simulated large
protein production in E. coli in a 2 l bioreactor scale conditions
Effect of glucose gradients Two fermenter system, stirred Biomass yield decreased about George et al. [35, 36]
in Baker's yeast production tank reactor (15 l) in series with 7% compared to ideal mixed
a tubular reactor (0.8 l); reference system. Gassing power
large-scale (215 m3) bubble column of produced yeast increased
10±25% compared to an ideal
mixed system
Effects of glucose gradients Large-scale (30 m3) and a scaled Highest level of stress in Schweder et al. [7]
on physiology of E. coli down reactor system stirred tank substrate entrance zone.
reactor (15 l) in series with a Transcriptional activation
tubular reactor (0.8 l). Monitoring occurred within about 10 s
of mRNA levels at different
positions
Effect of glucose gradients Two fermenter system STR+PFR Glucose-induced oxygen Bylund et al. [37]
in production of human (fed batch) Pilot scale (3 m3) limitation indicated by formate
growth hormone (hGH) formation. Cell lysis increased
in E. coli in SDR compared to well
mixed control experiments, but
degradation of rhGH monomer
decreased

There are several different mechanistic reasons behind How can the problems of gradients be avoided?
the effects of gradients. There will be direct mass action One method of avoiding potential problems is of course
effects if the cells are constitutively equipped for both to avoid creating the gradients. Use of multiple substrate
aerobic and anaerobic metabolism. Ethanol formation in inlets is here often preferable to increasing the stirrer rate.
S. cerevisiae will, for example, occur very rapidly if oxygen In the long run, it may also be worthwhile to use metabolic
is depleted. Also, the intracellular response in NADH levels engineering to decrease the sensitivity of the production
to a high glucose concentration has been found to occur organisms to the gradients. Tsai et al. [42] expressed
within a few seconds [40]. Furthermore, oxygen and glu- haemoglobin from Vitreoscilla in E. coli. The constructed
cose trigger several regulatory responses. As discussed by strains could thus ``store'' oxygen for use in times of ox-
Konz et al. [41], the regulatory response time will be de- ygen starvation and a decreased excretion of fermentation
termined by the rate of mRNA formation and the rate of products was indeed reported. Other studies have focused
translation. In a study by Schweder et al. [7], seven dif- on the over¯ow metabolism in E. coli. Reductions of ace-
ferent mRNA levels in E. coli were studied in a scaled- tate formation have been achieved by, e.g. heterologous
down system. It was found that the transcription of several expression of acetolactate synthase from Bacillus subtilis
genes, such as the proU gene, which is involved in os- [43], or by mutation of the pyruvate kinase or the phos-
moregulation, responded within 15 s of exposure to high photransferase system [44]. In the ®rst approach the ¯ux
glucose concentration. The rate of protein synthesis is to acetate was diverted into acetoin, whereas in the other
probably somewhat lower. The peptide elongation rate has strategy the maximum speci®c growth rate was reduced,
been estimated to be between 13±16 amino acids per thereby avoiding over¯ow metabolism.
second in E. coli [40]. For this reason, a single ``dip'' in
oxygen concentration of short duration may not result in Final remarks
the formation of a regulated protein. However, the effects It is certainly not simple to analyse the bioreactor. An
of repeated depletions occurring for a cell circulating in a increased usage of CFD modelling will be needed to
large-scale bioreactor are dif®cult to predict. obtain a better description of ¯ow patterns and
Bioprocess and Biosystems Engineering 24 (2002)

concentration gradients, and will help in the design of 19. De Lisa PM, Li J, Rao G, Weigand WA, Bentley W (1999) Moni-
less gradient-prone substrate inlet systems. However, toring GFP-operon fusion protein expression during high cell
density cultivation of Escherichia coli using an on-line optical
important problems regarding modelling of turbulence sensor. Biotechnol Bioeng 65:54±64
and multiphase ¯ow remain to be solved. Cellular phys- 20. Altenbach-Rehm J, Nell C, Arnold M, Weuster-Botz D (1999)
iology needs to be studied at process conditions, i.e. Parallel bubble columns with fed-batch technique for microbial
using high-density cultivations and industrial media. process development on a small scale. Chem Eng Technol
22:1051±1058
Tools developed in the ®eld of functional genomics must 21. Johnson A (1987) The control of fed-batch fermentation processes
be utilised to get an insight into responses on transcript, ± a survey. Automatica 23:691±705
protein and metabolite level, and the question of dynamic Ê kesson M, Hagander P, Axelsson JP (1999) A probing feeding
22. A
responses must be both theoretically and experimentally strategy for Escherichia coli cultures. Biotechnol Tech 13:523±
addressed. New approaches in kinetic modelling are 528
278 Ê kesson M, Nordberg-Karlsson E, Hagander P, Axelsson JP, Tocaj
23. A
required in which also the transcriptional and A (1999) On-line detection of acetate formation in Escherichia coli
translational levels are included. By combining novel cultures using dissolved oxygen responses to feed transients.
experimental and computational tools, it will certainly be Biotechnol Bioeng 64:590±598
possible to design smarter production systems for the 24. Taherzadeh M, Niklasson C, Liden G (2000) On-line control of
fed-batch fermentation of dilute-acid hydrolyzates. Biotechnol
biotechnology industry of tomorrow. Bioeng 69:330±338
25. Taherzadeh M, Gustafsson L, Niklasson C, Liden G (1999) Con-
References version of furfural in aerobic and anaerobic batch fermentation of
1. Cooney CL (1983) Bioreactors: design and operation. Science glucose by Saccharomyces cerevisiae. Biosci Bioeng 87:169±174
19:728±740 26. Guillard F, TraÈgaÊrdh C (1999) Modeling of the performance of
2. Feijen J, Hofmeester JJM, Groen D (1994) Scale-up of heteroge- industrial bioreactors with a dynamic microenvironmental ap-
neous bioprocesses. In: Alberghina L, Frontali L, Sensi P (eds) proach: a critical review. Chem Eng Technol 22:187±195
ECB6: Proceeedings of the 6th European Congress on biotech- 27. Steel R, Maxon D (1966) Dissolved oxygen measurements in pilot-
nology. Elsevier, Amsterdam, pp 919±926 and production-scale novobiocin fermentations. Biotechnol Bio-
3. Fiechter A, Fuhrmann GF, KaÈppeli O (1981) Regulation of glucose eng 8:97±108
metabolism in growing yeast cells. Adv Microbiol Physiol 22:123± 28. Manfredini R, Cavallera V, Marini L, Donati G (1983) Mixing and
183 oxygen transfer in conventional stirred fermentors. Biotechnol
4. Rose AH 1979 History and scienti®c basis of large-scale produc- Bioeng 25:3115±3131
tion of microbial biomass. In: Rose AH (ed) Economic microbi- 29. Bylund F, Guillard F, Enfors SE, TraÈgaÊrdh C, Larsson G (1999)
ology, vol 4. Academic Press, London, pp 1±29 Scale down of recombinant protein production; a comparative
5. Majewski RA, Domach MM (1990) Simple constrained-optimi- study of scaling performance. Bioproc Eng 20:377±389
zation view of acetat over¯ow in E. coli. Biotechnol Bioeng 30. Oosterhuis NMG, Kossen NWF, Olivier APC, Schenk ES (1985)
35:732±738 Scale-down and optimization studies of the gluconic acid fer-
6. Taherzadeh MJ, Niklasson C, Liden G (1999) Conversion of di- mentation by Gluconobacter oxydans. Biotechnol Bioeng 27:
lute-acid hydrolyzates of spruce and birch to ethanol by fed-batch 711±720
fermentation. Biores Technol 69:59±66 31. Sweere APJ, Mesters JR, Janse L, Luyben KCAM, Kossen NFW
7. Schweder T, Kruger E, Xu B, Jurgen B, Blomsten G, Enfors SE, (1988a) Experimental simulation of oxygen pro®les and their
Hecker M (1999) Monitoring of genes that respond to process- in¯uence on Baker's yeast production: I. One-fermentor system.
related stress in large-scale bioprocesses. Biotechnol Bioeng Biotechnol Bioeng 31:567±578
65:151±159 32. Sweere APJ, Janse L, Luyben KCAM, Kossen NFW (1988b) Ex-
8. Brune A (1998) Termite guts: the world's smallest bioreactors. perimental simulation of oxygen pro®les and their in¯uence on
Trends Biotechnol 16:16±21 Baker's yeast production: II. Two-fermentor system. Biotechnol
9. Sittig W (1982) The present state of fermentation reactors. J Chem Bioeng 31:579±586
Tech Biotechnol 32:47±58 33. Namdev PK, Thompson BG, Gray MR (1992) Effect of feed zone in
10. Kossen NWF 1985 Bioreactors: consolidation and innovation. fed-batch fermentations of Saccharomyces cerevisiae. Biotechnol
In: Proceedings of the 3rd European Congress on biotechnology, Bioeng 40:235±246
vol 4. VCH, Weinheim, pp 257±279 34. Namdev PK, Irwin N, Thompson BG, Gray MR (1993) Effect of
11. Nielsen J 1997 Physiological engineering aspects of Penicillium oxygen ¯uctuations on recombinant Escherichia coli fermenta-
chrysogenum. World Scienti®c, Singapore tion. Biotechnol Bioeng 41:666±670
12. Hempel DC, Dziallas H Scale-up, stirred-tank reactors. In: 35. George S, Larsson G, Enfors SO (1993) A scale-down two-com-
Flickinger MC, Drew SW (eds) 1999 Encyclopedia of bioprocess partment reactor with controlled substrate oscillations: metabolic
technology: fermentation, biocatalysis, bioseparation. Wiley, New response of Saccharomyces cerevisiae. Bioproc Eng 9:249±257
York, pp 2314±2332 36. George S, Larsson G, Olsson K, Enfors SE (1998) Comparison of
13. Jones RP, Green®eld PF (1982) Effect of carbon dioxide on yeast the Baker's yeast process performance in laboratory and pro-
growth and metabolism. Enzyme Microb Technol 4:210±223 duction scale. Bioproc Eng 18:135±142
14. Tanaka H (1981) Technological problems in cultivation of plant 37. Bylund F, Castan A, Mikkola R, Veide A, Larsson G (2000) In-
cells at high density. Biotechnol Bioeng 23:1203±1218 ¯uence of scale-up on the quality of recombinant human growth
15. Doran PM (1999) Design of mixing systems for plant cell sus- hormone. Biotechnol Bioeng 69:119±128
pensions in stirred reactors. Biotechnol Prog 15:319±335 38. Harrison DEF, Loveless JE (1971) Transient responses of facul-
16. Pickett AM, Topiwala HH, Bazin MJ 1979 A new method of in- tatively anaerobic bacteria growing in chemostat culture to a
dustrial bioreactor operation: the transient operation technique. change from anaerobic to aerobic conditions. J Gen Microbiol
Proc Biochem 14(11):10±16 68:45±52
17. Chae HJ, DeLisa MP, Cha HJ, Weigand WA, Rao G, Bentley WE 39. O'Beirne D, Hamer G (2000) Oxygen availability and growth of
(2000) Framework for online optimization of recombinant pro- Escherichia coli W3110: dynamic responses to limitation and
tein expression in high-cell-density Escherichia coli cultures using starvation. Bioproc Eng 23:381±387
GFP-fusion monitoring. Biotechnol Bioeng 69:275±285 40. Einsele A, Ristroph DL, Humphrey AE (1978) Mixing times and
18. Albano CR, Randers-Eichhorn L, Chang Q, Bentley WE, Rao G glucose uptake measured with a ¯uorometer. Biotechnol Bioeng
(1996) Quantitative measurement of green ¯uorescent protein 20:1487±1492
expression and chromophore cyclization. Biotechnol Tech 41. Konz JO, King J, Cooney CL (1998) Effects of oxygen on recom-
10:953±958 binant protein production. Biotechnol Prog 14:393±409
G. LideÂn: Understanding the bioreactor

42. Tsai PS, Hatzimanikatis V, Bailey J (1996) Effect of Vitreoscilla 44. Aristidou A, San KY, Bennett GN (1994) Modi®cation of
hemoglobin dosage on microaerobic Escherichia coli carbon and central metabolic pathway in Escherichia coli to reduce
energy metabolism. Biotechnol Bioeng 49:139±150 acetate accumulation by heterologous expression of the
43. Ponce E (1999) Effect of growth rate reduction and genetic Bacillussubtilis acetolactate synthase gene. Biotechnol Bioeng
modi®cations on acetate accumulation and biomass yield in 44:944±951
Escherichia coli. J Biosci Bioeng 87:775±780

279

You might also like