Metabolic Load of Recombinant Protein

Production: Inhibition of Cellular

Capacities for Glucose Uptake and
Respiration After Induction of a
Heterologous Gene in Escherichia coli
P. Neubauer,1 H. Y. Lin,2,3 B. Mathiszik3
1

Bioprocess Engineering Laboratory, P.O. Box 4300, Department of Process

and Environmental Engineering, Biocenter Oulu, University of Oulu,
FIN-90014 Oulu, Finland; telephone: +358 8 553 2384; fax: +358 8 553 2304;
e-mail: peter.neubauer@oulu.fi
2
Institut fur Genetik, Rheinische Friedrich-Wilhelms-Universitat Bonn,
Romerstrasse 164, D-53117 Bonn, Germany
3
Institut fur Biotechnologie, Fachbereich Biochemie/Biotechnologie,
Martin-Luther-Universitat Halle-Wittenberg, Kurt-Mothes-Str. 3,
D-06120 Halle, Germany
Received 15 August 2002; accepted 5 December 2002
DOI: 10.1002/bit.10645
Abstract: The strong expression of recombinant proteins
in bacteria affects the primary carbon and energy metabolism resulting in growth inhibition and acetate formation. By applying glucose pulses to fed-batch fermentations performed for production of a heterologous (glucosidase in Escherichia coli, we show that the
induction of the recombinant gene strongly inhibits the
maximum specific uptake capacities for glucose and the
respiration capacity. The accumulation of glucose in the
fermentation medium promotes the growth of plasmidfree cells. These inhibition effects are well described by
including the kinetics of product formation into a recently
published dynamic model (Lin et al. [2001] Biotechnol
Bioeng 73:349357). The new model also includes the
population characteristics and gives a good fit to the
measured data describing growth, production, substrate
consumption, by-product formation, and respiration.
2003 Wiley Periodicals, Inc. Biotechnol Bioeng 83: 5364,
2003.

Keywords: fed-batch fermentation; kinetic model; recombinant protein; acetate; population

INTRODUCTION
The microbial growth in fed-batch fermentations is mostly
controlled by limitation of the major carbon source to avoid
oxygen limitation and accumulation of toxic overflow metabolites. Therefore, the maximum uptake rate for the specific carbon source is a major parameter for the control of
the feed rate. Recently, we have shown that the maximum
uptake rates for glucose and oxygen in the Gram-negative
bacterium Escherichia coli are not constant throughout a
Correspondence to: Prof. Dr. Peter Neubauer
Contract grant sponsor: the Biotechnology program of the European
Community
Contract grant numbers: BIO-CT95-0028, BIO-CT98-0167

2003 Wiley Periodicals, Inc.

process, but depend on the specific growth rate (Lin et al.,

2001b). The results presented in this article indicate that in
glucose-limited fed-batch fermentations with constant feed
both parameters are strongly reduced during the feeding
phase, which has to be considered for the control of the feed
rate. Therefore, a specific variable for the uptake capacities
of glucose and oxygen was introduced into a structured
kinetic model, being defined as the maximum possible uptake rate at a certain time (see Lin et al., 2001b).
Here we apply this model to describe the performance of
fermentations with formation of a recombinant product. The
high formation of recombinant proteins competes with the
cellular synthesis and draws major resources to the product,
defined as the metabolic burden (Bentley et al., 1990). In the
past, many authors have referred to changes in the cell
physiology after strong induction of heterologous products,
such as induction of stress responses, growth inhibition, and
changes in the central carbon metabolism. Induction of recombinant products has been shown to increase the synthesis of acetate, which can finally lead to overaccumulation
and toxicity (San et al., 1994). A number of strategies have
been proposed to reduce the accumulation of acetate, such
as 1) selection or metabolic engineering of strains with a
low formation of acetate (e.g., Bauer et al., 1990; Aristidou
et al., 1994; San et al., 1994; Chou et al., 1994a,b) 2) use of
oxygen-enriched air or pure oxygen, 3) in situ removal of
acetate by perfusion systems, 4) use of alternative substrates
which reduce the formation of acetate, such as glycerol
(Anderson and von Meyenburg, 1980), and 5) coexpression
of hemoglobin to increase the oxidative capacity (Kallio et
al., 1994). Although all these possibilities are interesting
from an academic point of view, practically, in most recombinant processes the medium feed rate is adjusted in accor-

dance to the oxidative capacity of the cell population. However, this is not always easy, as the oxidative capacity is not
a constant since the critical specific glucose uptake rate
determining the formation of acetate varies.
Recently, we developed a pulse method, which is applied
in the following study for the determination of the uptake
capacities of glucose and oxygen in recombinant fed-batch
cultivations. In recombinant systems we always found a
reduction of these maximum uptake parameters after induction, which may result in the accumulation of glucose in the
culture broth. By simplifying this inhibition of the uptake
capacities for oxygen and glucose on the competition reaction between product formation and cellular syntheses, we
can incorporate product synthesis easily in our recently published kinetic model (Lin et al., 2001b) and succeed in a
fairly good description of the presented recombinant fedbatch fermentations.
MATERIALS AND METHODS
Strain
Escherichia coli RB791 (lacIqL8) is a derivative of W3110
[F, IN(rrnD-rrnE)1, ] and was kindly provided by the E.
coli Genetic Stock Center, Yale University (New Haven,
CT). The strain was transformed with the plasmid
pKK177glucC containing the -glucosidase gene (glucC)
of Saccharomyces cerevisiae. The expression of the glucC
gene is controlled by the isopropyl--D-thiogalactopyranoside
(IPTG) inducible tac-promoter (Kopetzki et al., 1989). Additionally, the strain carried the plasmid pUBS520, supplying the
minor argU tRNA at a constant higher level and additional
copies of lacI (Brinkmann et al., 1989).
Media and Culture Conditions
The fed-batch cultivations were performed in a 10-L Biostat
ED bioreactor (B. Braun Biotech, Allentown, PA) with an
initial culture volume of 4 L. All cultivations were carried
out in phosphate-buffered mineral salt medium with glucose
as carbon source and ammonia as nitrogen source and pH
control agent at a temperature of 35C as described in detail
by Teich et al. (1998). The cultivations were started with a
batch growth phase with an initial glucose concentration of
5 g L1. The addition of a 200 g L1 glucose containing feed
solution was started shortly before the initially added glucose was exhausted and was kept at a constant rate of 53.2
g h1. Induction was always performed by injection of 1
mM IPTG to the fermenter at 3 h after feeding start. The
cultivations were continued for approximately 20 h after
induction even if the product concentration was at the maximum a few hours after induction. The MFCS-Win program
(v. 1.1, B. Braun Biotech Int., Germany) was used for data
collection.
Analytical Methods
Cell Growth
Growth of the cultures was followed by measuring the optical density at 500 nm (OD500), by microscopic analysis

54

using a counting chamber and by determination of the cell

dry weight (CDW). Therefore, 1.5 mL of the cell suspensions were centrifuged in preweighed 2-mL test tubes and
washed once with 0.9% (w/v) NaCl solution. After removal
of the supernatant the samples were dried to constancy at
60C for at least 24 h. Colony-forming units were analyzed
by plating at least three dilutions of the culture broth and
each of these at triple on nutrient agar plates incubated for
13 days at 37C. The plasmid containing part of the population was calculated by replica plating from nonselective
plates to ampicillin-containing plates.

Product Concentration
The product concentration in relation to the total cellular
protein was calculated after separation of total cellular extracts by 12% SDS-PAGE and scanning of the appropriate
bands in correlation to samples of standards for -glucosidase, which were run in four concentrations on each gel.
The total protein concentration was determined by the Bradford assay with bovine serum albumin as reference protein.

Glucose and Oxygen Uptake Capacities

The glucose uptake capacity (qScap) was calculated from the
change of measured glucose to biomass concentrations during the batch phase and after induction, when the glucose
concentration was saturating the uptake systems according to
qScap=

1 . F
S + S Si .
V
X

During the glucose-limited growth phase a pulse technique

was applied, as described recently (Lin et al., 2001b).
Therefore, 0.51.0 g L1 pulses of glucose were added to
the cultivations; qScap was calculated from analyzed samples
which were frequently collected after the pulse or from the
DOT (dissolved oxygen tension) curve signal (Lin et al.,
2001b). The DOT was measured by a standard polarographic DOT electrode (Mettler-Toledo, no. 322336110).
The maximum oxygen uptake capacity was calculated from
the signal of the oxygen outgas analysis after a glucose
pulse.

Glucose and Acetate

Samples from the cultivation were directly chilled on ice
with further centrifugation (3 min, 13,000 rpm, 4C) and
collection of the supernatant or by direct filtration from the
reactor through a 0.2 m disk filter. Afterwards the samples
were stored at 20C for further analysis. The analysis of
glucose was performed with the hexokinase/glucose-6phosphate-dehydrogenase method (Kit no. 139106, Roche,
Germany) scaled down to microliter plates with four paral-

BIOTECHNOLOGY AND BIOENGINEERING, VOL. 83, NO. 1, JULY 5, 2003

lels for each sample as described by Larsson and Tornkvist

(1996). Acetate was analyzed with Kit no. 148261 (Roche,
Germany).
Outgas Analysis
The outgas oxygen and carbon dioxide concentrations were
measured with an URAS10E (Hartmann & Braun, Frankfurt, Germany). The specific rates (qO, qCO2) were calculated continuously over the fermentation by including the
biomass values fitted in intervals by polynomial functions
and the actual reactor volume, which was calculated from
the total reactor weight corrected for the broth density.
Adenosine Nucleotides
The adenosine nucleotides were analyzed according to an
HPLC method as described previously (Meyer et al., 1999).
Modeling and Computer Simulation
For simulation of the fermentations the SCILAB program
(INRIA, www.inria.fr) was used. The kinetic model applied
was described in detail earlier (Lin et al., 2001b) but was
modified to implement the changes of the metabolic flows
after induction of the recombinant proteins as described in
the Results section.
RESULTS AND DISCUSSION
Recombinant E. coli cells were grown in fed-batch cultivations with an initial batch phase and a following constantrate feeding strategy as described by Teich et al. (1998) and
Lin and Neubauer (2000). Induction was always performed
3 h after feeding start. The aim of this study was to investigate whether the growth inhibition following induction can
be explained as a result of the change of the glucose uptake
parameters. Therefore, the maximum uptake capacities for
glucose and oxygen were calculated from the DOT and
outgas signals following glucose injections in combination
to offline analyses of glucose and acetate by a recently
described method (Lin et al., 2001b). The data were used to
further develop the earlier published structured kinetic
model for recombinant systems with a phase shift (noninduction phase and induction phase). This model was further
extended to describe the appearance of different populations
after an induction event, thereby concentrating on the appearance of plasmid-free cells.
Effects of Induction on High Cell Density Growth
Data for biomass growth and product formation in cultivations with E. coli RB791 (pKK177glucC pUBS520) are
depicted in Figure 1. The measured values from three fermentations are well correlated. The cells grew always with
a maximum specific growth rate (max) of 0.79 0.05 h1

Figure 1. Fed-batch fermentation of E. coli RB791 (pKK177glucC

pUBS520) expressing a recombinant -glucosidase. The figure shows the
measured fermentation data (symbols) and the simulated values (lines).
The time of feed start (dashed line) and the time of induction (dotted line)
are indicated in each graph. The lowest graph shows the total cell count
(circles) and the viable cell counts estimated by replica plating for
pKK177glucC-containing cells (triangles) and pKK177glucC-free cells
(squares).

during the batch phase. Shortly before the initially added

glucose was exhausted the feed with a constant rate was
started. Deviations in the growth between the different cultures are mainly due to slight variations of the feeding start.
During the early fed-batch phase the glucose level was low
and consequently the acetate originating from the batch
phase was consumed. Three hours after feeding start, 1 mM
IPTG was added to induce the formation of -glucosidase.
The level of -glucosidase increased after induction by a

NEUBAUER ET AL.: INHIBITION OF GLUCOSE UPTAKE AND RESPIRATION

55

specific rate of 26 mg g1 h1 and reached a maximum of

about 400 mg L1 about 45 h after induction. From 2 h
after induction on, we detected an increasing glucose concentration in the cultivation broth. With a slight delay, acetate also appeared. Although the growth stopped shortly
after induction, the culture density increased again if the
cultivations were prolonged. This further increase in the
biomass concentration was due to overgrowth of a plasmidfree cell population. We were interested in studying this
phenomenon since plasmid-free cells had been detected before induction to be less than 1% of the whole culture.
The glucose uptake capacity of the culture was calculated
by evaluation of the DOT signal after 0.5 g L1 glucose
injections, as described previously (Lin et al., 2001b) or by
simple measurement of the glucose concentration during the
periods where glucose was not growth-limiting. The analysis showed that the uptake capacity for glucose was rapidly
declining after induction, but the measured signal increased
towards the end of the cultivation when plasmid-free cells
became a relevant population (see Fig. 2).
In the described system a fast and transient response was
also detected. IPTG addition caused increased respiration,
as is obvious from the specific oxygen uptake rate qO and
DOT (see Fig. 3). However, later qO decreased to a minimum of about 2.5 h after induction. During further cultivation, qO increased again transiently with a peak 8 h after
induction and a following shoulder. This behavior of qO is
in correlation with the ATP level and energy charge (Fig. 3),
which shows a peak shortly after induction as well a second
increase 8 h after induction.

Figure 2. Glucose uptake capacity qSmax during fed-batch fermentations

of E. coli RB791 (pKK177glucC pUBS520) with induction (filled symbols) and without induction (control, open symbols). Different symbols
refer to different methods for analysis of the glucose uptake capacity.
Triangles: calculation from the glucose concentration; circles: evaluation
of the DOT signal after saturating glucose pulses according to Lin et al.
(2001b); squares: analysis of the glucose concentration after saturating
glucose pulses. The time of feed start (dashed line) and the time of induction (dotted line) is indicated. The lines indicate the tendency of the values
during the fermentation time.

56

Figure 3. ATP concentration (), Energy charge (), qO, qCO2

(thin lines), and DOT (thick lines) during fed-batch cultivation of E. coli
(pKK177glucC pUBS520) expressing -glucosidase. The time of feed start
(interrupted line) and the time of induction (dotted line) are indicated in
each graph.

Structured Simulation Model for Recombinant

Fed-Batch Fermentations With Induction
The above-described complicated kinetics following induction are difficult to evaluate since -glucosidase induction
not only competes with cell metabolism and leads to growth
inhibition, but furthermore the decrease in glucose uptake
causes the growth of a cell population not containing the
pKK177glucC plasmid (see Fig. 1). To evaluate the impact
of the different cell populations on the fermentation data, we
looked for a model to describe these complex kinetics.
Therefore, we implemented simple product formation kinetics into our dynamic model (Lin et al., 2001b) and additionally considered a second, nonplasmid-containing population. The application of this model would also result in
more information about the behavior of the plasmidcontaining population following induction.
The applied model is based on nonconstant parameters
for the cellular uptake capacities for glucose and oxygen.
We earlier showed that these parameters (qOcap, qScap) are
dependent of the specific growth rate. Therefore, we introduced the parameters ES and EO, being the hypothetical
enzyme amount for substrate and oxygen uptake, respectively. The amount of ES and EO enzyme systems sets the
limits for the uptake capacities for glucose and oxygen (see
Fig. 4). ES and EO are synthesized from the carbon and
energy equivalents contributing to X. They decrease by their

BIOTECHNOLOGY AND BIOENGINEERING, VOL. 83, NO. 1, JULY 5, 2003

Table I.

Parameters and initial variables used in the simulations for giving the best fit.

Parameter/
variable

Unit

Fed-batch culture
of E. coli RB791
(pKK177glucC
pUBS520)

S0
V0
X0
CA
CS
CX
DOT*
F
H
KA
KLa
KS
qAcmax
qd
qEOd
qESd
qm
qOmax
qseg
qSmax
Si
sqP
tFStart
tind
YA/S
YES/X
YEO/X
YO/A
YO/S
YP
YX/A
YX/S,of
YX/S,ox

g L1
L
g L1
mol C g1
mol C g1
mol C g1
%
g h1
% L g1
g L1
h1
g L1
g g1 h1
g g1 h1
g g1 h1
g g1 h1
g g1 h1
g g1 h1
g g1 h1
g g1 h1
g L1
%
h
h
g g1
g g1
g g1
g g1
g g1
g g1
g g1
g g1
g g1

5
4
0.0001
0.0333
0.0333
0.04
100
0.053
14000
0.1
400
0.01
0.04
0
0.4
0.4
0.07
0.875
0.0004
2
220
80
0
3
0.667
0.07982
0.1
1.067
1.067
0.1
0.35085
0.15
0.51

Origin/source
Input concentration
Input concentration
Input concentration
Chemical formula (Xu et al., 1999)
Chemical formula (Xu et al., 1999)
Stoichiometric analysis (Xu et al., 1999)
Constant
Input concentration
Constant
Data fitting (Lin et al., 2001b)
Data fitting
Data fitting (Lin et al., 2001b)
Experimental data, data fitting
Experimental data: <0.005, (Lin and Neubauer, 2000)
Data fitting
Data fitting
Data fitting
Experimental data, data fitting
Data fitting
Experimental data, data fitting
Input concentration
Data fitting
Input concentration, start time of feeding
Input concentration, time of induction
Stoichiometric constant (Xu et al., 1999)
Data fitting
Data fitting (Lin et al., 2001b)
Stoichiometric constant (Xu et al., 1999)
Stoichiometric constant (Xu et al., 1999)
Data fitting
Experimental data, data fitting (Xu et al., 1999)
(Varma and Palsson, 1994; Chen et al., 1997)
Experimental data (Xu et al., 1999)

specific degradation rates qEOd and qESd and additionally by

death of X. Although the model ignores the complicated
glucose uptake kinetics in E. coli due to the phosphotransferase system and also the different pathways for glucose
uptake, we showed that fed-batch data are described more
thoroughly with the new model than with earlier fed-batch
models.
For the description of recombinant fermentations, we extend the recently published model (Lin et al., 2001b) regarding product (P) formation. In the model, we define P to
be a separate population beside the biomass X. This is important, since the enzymes EO and ES are part of the biomass
X, but are not produced by the product P. However, as the
product accumulates inside the cell, both X and P are added
for data presentation to give the total cell weight.
At the induction point, product formation is induced by
the chemical inductor IPTG directing a part of the metabolic
flow from the formation of X to the synthesis of P (Fig. 5a).
This product formation rate qP, has a defined maximum rate
only limited by the availability of carbon/energy determining the total rate for (see Fig. 5b). qP is only lowered if the
total available biomass synthesis capacity () is smaller

than qP; then is equal to qP and no biomass is formed.

Alternatively, one could assume that only a specific relative
part of the carbon flow to the biomass is used for qP (Fig.
5c).
In our model the metabolic load of product formation is
due to the reduced formation of ES and EO, and by this
affects the further cellular syntheses. Although by this simplification we reduce all metabolic activities to the available
substrate and energy equivalents created by respiration, we
obtain a reliable fit of our experimental values (see Fig. 1).
The new kinetic model can now simulate the measured
increase of glucose after induction of the recombinant product. However, in E. coli RB791 (pKK177glucC pUBS520)
the available glucose is consumed by the small population
of plasmid-free cells, allowing them to grow exponentially,
with maximum growth rate to a significant population. To
simulate the levels for glucose, acetate, and the respiration
pattern we included a second cell population into our model,
which does not produce the -glucosidase product and
therefore is not affected by IPTG (see Fig. 6).
It should be noted that only the plasmid pKK177glucC is
lost by segregation, but the second plasmid pUBS520 is

NEUBAUER ET AL.: INHIBITION OF GLUCOSE UPTAKE AND RESPIRATION

57

Figure 4. Simplified scheme of the simulation model describing the metabolic flow rates applied for the description of recombinant fed-batch processes.
The figure is a schematic representation of the model presented by Lin et al. (2001b) and additionally includes the product formation kinetics.

stably maintained in all cells. No cells without this plasmid

were detected in the cultivations (data not shown). The
death rate constant qd for the bacteria was first considered in
the model; however, qd was so small (qd < 0.005) that it
could be neglected for the fed-batch fermentation of E. coli
(pKK177glucC pUBS520) (for data, see Lin and Neubauer,
2000).

Figure 5. Principle of substrate flow before and after induction. (a) Simplified flow of carbon/energy substrate to either support biomass growth,
maintenance or product formation before induction (left) and after induction (right). (b) and (c) show two different models of the competition effect
between product (P) formation (dark area) and active cell (X) production.
(b) qP is induced to a specific rate which is only limited by the available
resources; (c) qP is directing a constant share of the carbon/energy resources from X to P.

58

Model Application
First, it was tested whether the model is sensitive to the two
different variants for qP, either defining it with a maximum
rate lowered only if the available cellular resources decrease
below the limit to allow this maximum rate (Fig. 5b, case
A), or by defining that qP can only occupy a specific part of
the total resources, which is constant over time and determined by the strength of induction (Fig. 5c, case B). In a
simulation where only this part of the model is varied but

Figure 6. Population characteristics in the -glucosidase fed-batch fermentation. The plasmid-containing populations X1 and X1P can segregate
with qseg into plasmid-free cells X2. All cell populations die by the death
rate constant qd.

BIOTECHNOLOGY AND BIOENGINEERING, VOL. 83, NO. 1, JULY 5, 2003

the initial rate qP at the point of induction and all other

parameters are the same, only small differences can be seen
in the calculated variables (see Fig. 7). Surprisingly, no
significant differences can be detected among the curves for
total biomass and the subpopulations or in the product and
acetate concentrations. However, case A leads to a stronger
inhibition of the synthesis of EO and ES and due to the lower
content of these enzymes glucose accumulates faster and to
higher levels (Fig. 7b). This higher level of glucose causing
qS to increase according to MONOD kinetics and correspondingly higher qO leads to a sharper decrease of qO if
glucose gets limited again by the growth of the plasmid-free
population (Fig. 7c). In contrast to case A, the decrease of
qO and the decrease of S at this phase are rather smooth in
case B.
In Figure 8 the parameters qO and qS are separately calculated for the two cell populations X1 and X2, with and
without the pKK177glucC plasmid. This figure illustrates
the behavior of qScap and qOcap. For both populations qScap
and qOcap decrease during the fed-batch phase, but qS and qO
decrease below the capacity. This leaves free respiratory
capacity for the uptake of acetate (see Figs. 1, 6). However,
after induction the response of the two populations is differentqScap and qOcap are rapidly decreased in X1 due to
the competitive influence of product formation and 3 h after
induction qS and qO are limited by the capacity of the corresponding uptake systems. In contrast, both parameters
qScap and qOcap are increased in the plasmid-free population
X2 after induction due to the available glucose in the cultivation medium. At 8 h after induction, when glucose
reaches the growth-limiting concentration, qS and qO decrease again, but remain lower than the corresponding capacity of the uptake system.
Although we have no continuous measurements for qS,

the values for qScap correspond well to the measured values

presented in Figure 2. Also, the continuous calculation of qO
from the outgas oxygen values reflects the principle shape
of the simulated qO curve. Of course, the model is not able
to simulate the transiently increased respiration after induction according to Figure 3. A further limitation of the model
can be observed at 10 h after induction. As discussed above,
the model does not consider similar transient changes of the
acetate consumption and uptake systems as it does for carbon/energy substrate and oxygen. Therefore, the second
smaller peak in the measured qO curve, which is due to the
consumption of acetate, does not appear in the simulation
of qO.
The basis of the adapted model is the competition of
product formation for cellular precursors and energy withdrawn from the formation of reproducible biomass. Shortly
after induction the initial part of -glucosidase synthesis on
the total translation can be roughly estimated from the specific rate for -glucosidase synthesis (qP) of 0.26 g g1 h1,
the formation of protein, which is approximately 55% of the
cell dry mass (YProtein/X) (Neidhardt and Umbarger, 1996),
and the relative translation activity at the point of induction
determined earlier to be about 71% of the activity detected
at the maximum specific growth rate during the batch phase
(Lin et al., 2001a). The total protein synthesis rate, therefore, is about 0.31 g g1 h1 at the point of induction, being
71% of the maximum protein synthesis rate of 0.43 g g1
h1 determined during the batch phase. This approximation
estimates the initial -glucosidase synthesis to be about
84% of the total protein synthesis in the cell and more than
45% of biomass synthesis from carbon substrate.
This metabolic flow is considered in the model by sqP.
The potential influence of sqP on the fermentation is evaluated in Figure 9. Therefore, using the case A model, sqP was

Figure 7. Simulations of the -glucosidase process by use of the two different models for the synthesis of product. Model where qP is only limited by
according to Figure 5b () and model where qP is defined with a certain part of according to Figure 5c (----).

NEUBAUER ET AL.: INHIBITION OF GLUCOSE UPTAKE AND RESPIRATION

59

Figure 8. Simulated values for the substrate uptake rates qS and qScap (a) and for the respiration rates qO and qOcap (c) of the two populations X1 and X2.
Graph (b) shows the total substrate uptake rates qS and qScap for the total biomass (X1 + X2). Correspondingly, the total oxygen uptake rates qO and qOcap
are shown in graph (d). The qO values calculated from on-line oxygen analysis in the outgas air are shown by open circles.

varied whereas all other parameters were kept constant. Interestingly, the strength of induction with sqP of 0.8 gave
the best fit to the measured values. A higher ratio of product
synthesis did not affect the curves significantly, indicating
the strong competitive effect of a high product synthesis on
the cellular system. However, if sqP was decreased the
maximum product concentration obtained in the process remained approximately constant, but the rate of product accumulation decreased. Also, due to the lower inhibitory effect on the EO and ES systems at lower sqP values, glucose
accumulation and growth of plasmid-free cells were significantly delayed.
CONCLUSIONS
In this study we show that the overexpression of -glucosidase strongly inhibits the maximum respiration activity
and the maximum glucose uptake rate of the host cells. We
consider this to be a consequence from the competition for
cellular resources of the strong synthesis of the -glucosidase product resulting in a lower synthesis of cellular housekeeping proteins. The presented data complete the picture of
the physiological reactions in response to -glucosidase induction, which are characterized by an inhibition of cellular
processes such as chromosomal replication, transcription,
and translation as well as by noninduction of cellular stress
responses such as stringent response, general starvation response, and SOS response (see Teich et al., 1998; Lin et al.,
2001a).

60

Based on the experimental data, we developed a structured kinetic model considering the transient change of the
uptake of glucose and oxygen in dependence on the synthesis of the hypothetical uptake complexes. By applying the
concept of competition of product formation for cellular
precursors and energy, we succeed in fitting the experimental data for the different cell populations, glucose, acetate,
specific oxygen uptake, and product formation.
We measured the uptake capacity for glucose and oxygen
in fed-batch fermentations with other strongly induced recombinant products and found an inhibition of the capacities for glucose uptake and respiration also in these other
systems, although to a different degree (data not shown).
Specifically, the inhibiting effect on qScap and qOcap was
lower for recombinant products which did not inhibit the
growth. One of these systems was investigated for expression of the general starvation response regulator S and we
found it to be induced (Lin, 2000). In contrast, S was
decreasing after induction of -glucosidase (Schweder et
al., 2002). Therefore, we suggest that our way of simply
structuring the metabolic precursors into the flow for product and cell mass is only valid for systems where the metabolic flow is continuing after induction without being controlled by additional regulatory systems of the cell. However, in recombinant systems where the metabolic flows are
controlled by induction of responses such as the stringent
response and the general starvation response, we assume an
additional change of the parameters YES/X and YEO/X. It
should be remarked that the model does not include the fast

BIOTECHNOLOGY AND BIOENGINEERING, VOL. 83, NO. 1, JULY 5, 2003

of data which would be included in metabolic flow models.

Furthermore, it is not possible to evaluate the specific limiting steps within the enzyme cascades directly from the
model. However, the data and the model suggest that the
activation of the respiratory chain or the use of alternative
carbon substrates for induction might be promising strategies to improve the yield of recombinant proteins. In principle, this has been practically shown by increased product
yields when hemoglobin was co-overexpressed (Tsai et al.,
1996a,b) or when three carbon sources glucose, glycerol,
and lactose were added simultaneously during fed-batch
cultivations (Viitanen et al., unpub. data).
The inhibition of the glucose uptake of the productforming recombinant cells is interesting and relevant from
the practical point that the overgrowth of plasmid-free cells
is exclusively related to it. Therefore, strategies to either
control the feed in a way to keep the glucose low after
induction or to avoid the decrease of the glucose uptake rate
could have a beneficial effect on the total process yield. This
is especially interesting for recombinant products which are
accumulated over a long time interval, such as correctly
folded disulfide bond-containing proteins.
Independently of the limitations of a structured kinetic
model in general, our new model allows a good fit to the
experimental data as presented here for one recombinant
system and should be a valuable tool for the determination
of the effect of changes in process performance, such as
changes in the feed rate or in the strength of induction.
The authors thank the E. coli Stock Center (New Haven, CT,
USA) for strain RB791 and Roche Diagnostics for the plasmids
pKK177glucC and pUBS520.

APPENDIX

Simulation Model

Figure 9. Test of parameter dependencies for the described structured

kinetic model. Variation of sqP: 0.2----, 0.4 , 0.6 ........., 0.8 ,
0.9 - - -.

To describe the development of product formation and the

segregation of plasmid and plasmid-free cells, we have extended the model of Lin et al. (2001b). Here we show the
differential equation system and the calculation of the initial
values. The formulas for calculation of all subsidiary variables have been presented in the earlier article. However,
the kinetic parameters must be calculated for both populations, X1 and X2.

Differential Equation System

metabolic changes within the minutes time window after
induction, such as the fast and transient increase of respiration indicated in Figure 3. This is because the model does
not consider physiological changes due to gene regulation
by stress responses nor does it include the extensive series

V = F

[1]

F
X1 = + 1 qseg qd X1
V

NEUBAUER ET AL.: INHIBITION OF GLUCOSE UPTAKE AND RESPIRATION

[2]

61

F
X2 = + 2 qd X2 + qseg X1
V

[3]

F
S = S Si qS1 X1 qS2 X2
V

[4]

Calculation of Subsidiary Variables

The following equations complete the model. All variables
must be calculated twice, for X1P and X2. For better readability, the indexes 1P and 2 are omitted.

T = K a DOT* DOT q X + q X H
DO
L
O1
1
O2
2
[5]
F
A = A + qAp1 qAc1 X1 + qAp2 qAc2 X2
V

qScap = qS max

rS
rS0

qOcap = qO max

rO
rO0

[6]

qS = qScap

S
S + KS

rS1 = YES X 1 rS1 qESd + 1

[7]

rS2 = YES X 2 rS2 qESd + 2

[8]

The real maintenance coefficient qmr cannot be greater than

qSox, therefore qmr min{qm, qSox}.

rO1 = YEO X 1 rO1 qEOd + 1

[9]

qSoxan = qSox qmr YX Sax CS

rO2 = YEO X 2 rO2 qEOd + 2

[10]

F
P = P + qP X1
V

qSox = qS

qSoxen = qSex qSoxan

qOx = qSoxen YO S

[11]

We assume that an enzyme E (i.e., EO or ES) grows as a

constant part r of the growing biomass X and becomes inactive by a first-order kinetics with degradation rate qEd.
Thus:
E = YE X X r X qEd

If the calculated value for qOs is greater than qOcap, we set

qOs = qOcap and correct the values of qSoxan, qSoxen, qSox, and
qmr:
qSoxen =
qSox =

qOcap
YO S
qmr CS YX Sox qSoxen
CS YX Sox 1

qSoxan = qSox qSoxen

By differentiation we obtain Eqs. [7] or [9] which describe

the inhibition of ES and EO for the different cell populations.
If max, then rS1, rS2, rO1, and rO2 should be zero.
Therefore, the initial values for rS1, rS2, rO1, and rO2 are:

If now qSox < qmr then we must set qmr qSox and therefore
we obtain qmr = qSox = qSoxen and qSoxan = 0.

1max
,
qESd + 1max
2max
,
rS20 = YES / X
qESd + 2max
1max
,
rO10 = YEO X
qEOd + 1max

qAc = qAc max

qSof = qS qSox

rS10 = YES X

2max
where
qEOd + 2max
i max = qSi max qm YX Sox i = 1, 2.
rO20 = YEO X

qSofan = qSof YX Sof CS

qSofen = qSof qSofan
qAcan = qAc YX A CA
qAcen = qAc qAcan
If qAcen > (qOcap qOS)/ YO / A then we set qAcen = (qOcap
qOS)/YO / A and we obtain:

Before induction, there is no product formation. Thus, qP 0.

At the moment of induction, the maximum specific product
formation rate is determined by qP max 1 YX / P sqP.
Let 1P the specific growth rate of population X1 including
the product, then after induction applies 1P 1 + qP YX/P.
If 1P YX / P < qP max then only the product grows, thus
qP 1P YX / P and 1 = 0. Otherwise is qP = qP max and
1 1P 1/YX / P qP (see Fig. 5B).

62

A
A + KA

BIOTECHNOLOGY AND BIOENGINEERING, VOL. 83, NO. 1, JULY 5, 2003

qAc =

qacen
1 y X A CA

qAcan = qAc qAcen

qAp = qSofen YA S
qO = qOs + qAcen YO A
= qSox qm YX Sox
+ qSof YX Sof qAc YX A

NOMENCLATURE

A
ATP
DOT
ES, EO
F
H
K
KLa
P
qd
qESd,qEOd
qA
qO
qP
qS
qseg
rO,rS
S
Si
sqP
T
V
X
X1
X1P
X2
Y
YES/X,YEO/X

specific growth rate [h1]

acetate concentration [g L1]
adenosine 5-triphosphate
dissolved oxygen tension [%]
concentration of the enzyme(s) for substrate or oxygen consumption [g L1]
feed flow rate [L h1]
constant derived from the Henrys law [14,000% L g1]
saturation constant [g L1]
volumetric oxygen transfer coefficient [h1]
product
specific death rate [g g1 h1]
specific death rates for the uptake enzymes of oxygen and
substrate [g g1 h1]
specific acetate production of consumption rate [g g1 h1]
specific oxygen uptake rate [g g1 h1]
specific product formation rate [g g1 h1]
specific substrate (glucose) uptake rate [g g1 h1]
specific segregation rate
ratio of enzyme for oxygen or substrate consumption per
biomass
substrate (glucose) concentration [g L1]
substrate concentration in the inlet feed solution [g L1]
induction strength related initial relative value for qP from
total synthesis capacity [relative share from 1]
cultivation time [h]
fermenter volume [L]
biomass concentration [g L1]
plasmid containing part of the biomass [g L1]
plasmid containing part of the biomass with product [g L1]
plasmid-free part of the biomass [g L1]
yield [g g1]
yield of the enzyme for substrate uptake or respiration from
the biomass growth

Subscripts
0
1, 2
A
an
c
cap
CO2
d
E
en
i
m
max
O
of
ox
P
S
X

Initial concentration
populations 1 (with plasmid) or 2 (without plasmid)
acetate
anabolic
consumption
capacity
carbon dioxide
degradation
enzyme
energetic
inlet concentration
maintenance
maximum
oxygen
overflow
oxidative
product
substrate (glucose)
biomass

References
Anderson KB, von Meyenburg K. 1980. Are growth rates of Escherichia
coli in batch cultures limited by respiration? J Bacteriol 144:114123.
Aristidou AA, San K-Y, Bennett GN. 1994. Modification of central metabolic pathway in Escherichia coli to reduce acetate accumulation by
heterologous expression of the Bacillus subtilis acetolactate synthase
gene. Biotechnol Bioeng 44:944951.

Bauer KA, Ben-bassat A, Dawson M, De la Puente VT, Neway JO. 1990.

Improved expression of human interleukin-2 in high-cell-density fermentor cultures of Escherichia coli K-12 by a phosphotransacetylase
mutant. Appl Environ Microbiol 56:12961302.
Bentley WE, Mirjalili N, Andersen DC, Davis RH, Kompala DS. 1990.
Plasmid encoded protein: the principal factor in the metabolic burden associated with recombinant bacteria. Biotechnol Bioeng 35:
668681.
Brinkmann U, Mattes RE, Buckel P. 1989. High-level expression of recombinant genes in Escherichia coli is dependent on the availability of
the dnaY gene product. Gene 85:109114.
Chen R, Yap WMGJ, Postma PW, Bailey JE. 1997. Comparative studies of
Escherichia coli strains using different glucose uptake systems: metabolism and energetics. Biotechnol Bioeng 56:583590.
Chou C-H, Bennett GN, San K-Y. 1994a. Effect of modified glucose
uptake using genetic engineering techniques on high-level recombinant protein production in dense Escherichia coli cultures. Biotechnol
Bioeng 44:952960.
Chou C-H, Bennett GN, San K-Y. 1994b. Effect of modulated glucose
uptake on high-level recombinant protein production in a dense Escherichia coli culture. Biotechnol Prog 10:644647.
Kallio PT, Kim DJ, Tsai PS, Bailey JE. 1994. Intracellular expression of
Vitreoscilla hemoglobin alters Escherichia coli energy metabolism
under oxygen-limited conditions. Eur J Biochem 219:201208.
Kopetzki E, Schumacher G, Buckel P. 1989. Control of formation of active
soluble or inactive insoluble bakers yeast alpha-glucosidase PI in
Escherichia coli by induction and growth conditions. Mol Gen Genet
216:149155.
Larsson G, Tornkvist M. 1996. Rapid sampling, cell inactivation and
evaluation of low extracellular glucose concentrations during fedbatch cultivation. J Biotechnol 49:6982.
Lin HY. 2000. Cellular responses to the induction of recombinant genes in
Escherichia coli fed-batch cultures. PhD thesis, Martin-LutherUniversity Halle-Wittenberg.
Lin HY, Neubauer P. 2000. Influence of controlled glucose oscillations on
a fed-batch process of recombinant Escherichia coli. J Biotechnol
79:2737.
Lin HY, Hanschke R, Nicklisch S, Nietsche T, Jarchow R, Schwahn C,
Riemschneider S, Meyer S, Gupta A, Hecker M, Neubauer P. 2001a.
Cellular responses to strong overexpression of recombinant genes in
Escherichia coli DNA relaxation and cell death after induction of
-glucosidase. In: Merten OW, Mattanovich D, Lang C, Larsson G,
Neubauer P, Porro D, Postma PW, Teixeira de Mattos J, Cole J,
editors. Recombinant protein production with prokaryotic and eukaryotic cells. A comparative view on host physiology. Dortrecht, The
Netherlands: Kluwer. p 5574.
Lin HY, Mathiszik B, Xu B, Enfors S-O, Neubauer P. 2001b. Determination of the maximum specific uptake capacities for glucose and oxygen
in glucose limited fed-batch cultivations of Escherichia coli. Biotechnol Bioeng 73:347357.
Meyer S, Noissomit-Rizzi N, Reuss M, Neubauer P. 1999. Optimized
analysis of intracellular adenosine and guanosine nucleotides in Escherichia coli. Anal Biochem 271:4352.
Neidhardt FC, Umbarger BE. 1996. Chemical composition of Escherichia
coli. In: Neidhardt FC, Curtiss III R, Ingraham JL, Lin ECC, Low KB,
Magasanik B, Reznikoff WS, Riley M, Schaechter M, Umbarger HE,
editors. Escherichia coli and Salmonella typhimurium cellular and molecular biology. 1316. Washington, DC: American Society of Microbiology.
San KY, Bennett GN, Aristidou AA, Chou CH. 1994. Strategies in highlevel expression of recombinant protein in Escherichia coli. Ann NY
Acad Sci 721:257267.
Schweder T, Lin HY, Jurgen B, Breitenstein A, Riemschneider S, Khalameyzer V, Gupta A, Buettner K, Neubauer P. 2002. Role of the

NEUBAUER ET AL.: INHIBITION OF GLUCOSE UPTAKE AND RESPIRATION

63

general stress response during strong overexpression of a heterologous

gene in Escherichia coli. Appl Microbiol Biotechnol 58:330337.
Teich A, Lin HY, Andersson L, Meyer S, Neubauer P. 1998. Amplification
of ColE1 related plasmids in recombinant cultures of Escherichia coli
after IPTG induction. J Biotechnol 64:197210.
Tsai PS, Hatzimanikatis V, Bailey JE. 1996a. Effect of Vitreoscilla hemoglobin dosage on microaerobic Escherichia coli carbon and energy
metabolism. Biotechnol Bioeng 48:139150.
Tsai PS, Nageli M, Bailey JE. 1996b. Intracellular expression of Vitreos-

64

cilla hemoglobin modifies microaerobic Escherichia coli metabolism

through elevated concentration and specific activity of cytochrome o.
Biotechnol Bioeng 48:151160.
Varma A, Palsson BO. 1994. Stoichiometric flux balance models quantitatively predict growth and metabolic by-product secretion in wildtype Escherichia coli W3110. Appl Environ Microbiol 60:37243731.
Xu B, Jahic M, Enfors SO. 1999. Modeling of overflow metabolism in
batch and fed-batch cultures of Escherichia coli. Biotechnol Prog 15:
8190.

BIOTECHNOLOGY AND BIOENGINEERING, VOL. 83, NO. 1, JULY 5, 2003