Biotechnol. Prog.

1997, 13, 249257 249

Enhanced Production of Human Mini-Proinsulin in Fed-Batch

Cultures at High Cell Density of Escherichia coli
BL21(DE3)[pET-3aT2M2]
Chul Soo Shin, Min Seon Hong, Cheon Soon Bae, and Jeewon Lee*
Bioprocess Engineering Laboratory, Hanhyo Institutes of Technology #461-6, Jeonmin-Dong, Yusong-Ku,
Taejon, South Korea

Synthesis of recombinant protein (human mini-proinsulin) is investigated in fed-batch

cultures at high cell concentration of recombinant Escherichia coli BL21(DE3)[pET-
3aT2M2]. Transcription of the recombinant gene is controlled by a T7 promoter
system. The human mini-proinsulin is characterized by a C-chain peptide consisting
of only nine amino acids, whereas the C-chain peptide of natural human proinsulin is
made up of 35 amino acids. It is expressed in a fusion protein with a small fusion
partner (a peptide with 18 amino acids) and finally aggregated into insoluble inclusion
bodies in cytoplasm of recombinant E. coli. The fermentative production of this small
fusion mini-proinsulin may be of great advantage in enhancing the yield of human
insulin. To find an optimum induction strategy, effects of various key cultivation
variables on the mini-proinsulin production are examined in high cell density fed-
batch cultures. No general correlation is found between preinduction specific growth
rate and recombinant protein synthesis, which confers a flexibility in choosing the
feeding strategy of preinduction media for achieving the high cell density cultures. A
culture temperature below 37 C is unfavorable for recombinant gene expression, and
the T7-based expression system is almost completely repressed at 30 C. The nutrient
glucose and yeast extract concentration in postinduction feed media is optimized by
applying a statistical method for medium optimization, i.e. response surface methodol-
ogy, and an effective amount of inducer molecule (IPTG) is determined to maximize
the specific recombinant protein formation. The mini-proinsulin production in E. coli
culture is significantly influenced by the volumetric feed rate of postinduction media,
which is shown to be closely related to the plasmid copy number in the recombinant
cell. Consequently, in a single-stage fed-batch process, the mini-proinsulin concentra-
tion is increased up to 7 g/L, approximately 62 wt % of which corresponds to mature
human insulin. A two-stage fed-batch fermentation process, with recombinant cell
growth occurring at a constant growth rate and constant cell concentration in a growth
fermenter and mini-proinsulin production in an induction fermenter, is designed, and
its efficacy in increasing volumetric productivity of mini-proinsulin is demonstrated.

Introduction ation (refolding), sulfitolysis, dialysis, ion-exchange chro-

The utilization of recombinant DNA (rDNA) technology
and key bioprocess technologies (e.g., fermentation, pro-
tein recovery and purification, etc.) has led to important
advances in the efficient and controlled production of
many eukaryotic proteins/peptides in foreign hosts such
as Escherichia coli. There are various factors that may
influence the synthesis of a particular protein from
recombinant E. coli: host-vector interaction, plasmid
stability, induction efficiency of promoter system(s),
protein stability, biomass concentration, and so on (Yee
and Blanch, 1992; Zabriskie and Arcuri, 1986; Georgiou,
1988). Sometimes protein stability is a major problem
in the production of small proteins or peptides such as
human proinsulin because of their short half-life in the
host cell (Talmadge and Gilbert, 1982). In order to
increase the stability of expressed protein in E. coli,
expression in fusion proteins has been widely employed.
Bacterial intracellular fusion systems are frequently
characterized by production of insoluble inclusion bodies
(Williams et al., 1982) which require complex down-
stream process, i.e. in vitro denaturation and renatur-
* To whom all correspondence should be addressed.
S8756-7938(97)00018-0 CCC: $14.00 1997 American Chemical Society and American Institute of Chemical Engineers
matography, reversed-phase HPLC, etc. (Marston, 1986;
Chen et al., 1995; Petrides et al., 1995; Kroeff et al., 1989).
It has been shown that protein recovery in refolding step
is sometimes low enough to significantly decrease the
overall recovery efficiency and that some protein loss is
inevitable in each recovery/purification step. Also, it is
time-consuming to find the optimal operating conditions
of such complex downstream process. It is advantageous
if the concentration and/or productivity of recombinant
protein is enhanced in the fermentation broth.
The importance of well-optimized fermentation process
may be significant in production of human proinsulin
from recombinant E. coli cultures. Human insulin was
the first marketed human health-care product derived
from rDNA technology. After its commercialization by
Eli Lilly and Co. in the U.S. in 1982, human insulin has
gradually replaced animal insulin as the most frequently
chosen treatment of insulin-dependent and insulin-
requiring diabetes and now accounts for about 70% of
the insulin used in the U.S. (Raskin and Clements, 1991).
As the selling price of human insulin is now very low
(less than $500/g) compared to other recombinant phar-
maceutical proteins in market (e.g., $850 000/g of Eryth-
250
ropoietin and $450 000/g of G-CSF), lowering the pro-
duction cost to a certain level is a critical requirement
for the commercial manufacture of human insulin. The
bulk market of human insulin requires large-scale fer-
mentation producing human proinsulin, and hence even
small improvements in protein concentration signifi-
cantly affect productivity of the fermentation process. The
downstream process for human proinsulin synthesized
as a cytoplasmic insoluble form in recombinant E. coli is
quite complex, and from the manufacturing standpoint,
it is therefore desirable to increase the concentration or
volumetric productivity of human proinsulin in the
fermentation process.
The present investigation is concerned with synthesis
of an insoluble recombinant protein, fusion human pro-
insulin in fed-batch cultures of E. coli BL21(DE3)[pET-
3aT2M2]. The fusion human proinsulin contains rela-
tively small fusion partner (18 amino acids), and the
C-chain peptide (35 amino acids) of natural human
proinsulin has been replaced by a new peptide sequence
consisting of only nine amino acids (Shin et al., 1995).
The reduced size of fusion proinsulin and hence the
increased purity of human insulin in insoluble recombi-
nant product may contribute toward increasing the yield
of mature human insulin in fermentation. This mutant
human proinsulin is named mini-proinsulin in this
paper. In the present work, the effects of key fermenta-
tion variables such as growth and production medium
compositions, preinduction specific growth rate, induction
temperature, inducer concentration, and volumetric feed
rate of postinduction media are systematically investi-
gated to find optimal conditions maximizing the biomass
production and the yield of recombinant mini-proinsulin.
Emphasis is also placed on the efficacy of two-stage, cyclic
fed-batch operation in obtaining increased productivity
of the recombinant protein. The two-stage, cyclic fed-
batch process contains aspects of both fed-batch and
continuous processes and, therefore, is advantageous in
increasing the productivity of recombinant products. It
is also very useful when optimal growth conditions are
not conducive for optimal product formation, which is
frequently encountered in production of recombinant
proteins from genetically-engineered microorganisms
containing inducible promoter systems (Lee and Parule-
kar, 1993, 1996; Parulekar and Lee, 1993).
Materials and Methods
E. coli Strain and Recombinant Plasmid Con-
struction. The host-plasmid system used in this study
consists of the structural gene for a fusion mini-pro-
insulin (Ins) carried on the recombinant T7-based pET
expression plasmid, pET-3aT2M2 (Amp+Chl+Ins+), with
the mini-proinsulin negative E. coli strain BL21(DE3)
(F-ompTrB-mB-) as the host. In the plasmid vector
pET-3a (4640bp), the T7 transcription/expression region
was excised as a 522bp HindIII-NdeI fragment (Figure
1). The host, BL21(DE3), contains a single chromosomal
copy of T7 RNA polymerase gene, controlled by lacUV5
promoter. The human mini-proinsulin has been pre-
pared by replacing the C-chain peptide of natural human
proinsulin with a small peptide which has strong -turn
structure, consisting of following amino acid sequence:
Arg-Arg-Tyr-Pro-Gly-Asn-Val-Lys-Arg (Shin et al., 1995).
To express the mini-proinsulin in a fusion protein, the
following peptide of 18 amino acids has been used as a
fusion partner: Pro-Ser-Asp-Lys-Pro-(His)10-Ser-Ser-Met
(first five amino acids corresponding to N-terminus 8th
to 12th amino acids of tumor necrosis factor (TNF) and
the last methionine being the cleavage site for cyanogen
bromide (CNBr)) (Shin et al., 1994).
Biotechnol. Prog., 1997, Vol. 13, No. 3
Figure 1. Recombinant plasmid construction: Ap, ampicillin
resistance gene; Ori, origin of replication; P/T7, T7 promoter;
T/T7, T7 terminator; NdeI, HindIII, restriction enzyme sites.
Media and Cultivation. Stock cultures, stored at
-80 C, were prepared by growing E. coli BL21(DE3)-
[pET3a-T2M2] cells to midexponential growth phase in
Luria Broth (LB) containing 200 mg of ampicillin/L,
followed by harvesting the cells and resuspending them
in a mixture of LB medium and glycerol (15%) containing
200 mg of ampicillin/L. The inoculum for each bioreactor
experiment was prepared by serial subculturing of re-
combinant cells taken from stock cultures using LB
medium containing 200 mg of ampicillin/L. The semi-
synthetic medium used for initial batch cultures con-
tained, per liter, (a) 5 g of KH2PO4, 3 g of K2HPO4, 1.67
g of (NH4)2SO4, 0.1 g of FeSO4H2O, 0.02 g of CaCl22H2O,
80 mL of trace metal solution, 0.1 g of thiamine-HCl, 20
g of yeast extract; (b) 2 g of MgSO47H2O; and (c) 20 g
of glucose. Components a, b, and c were autoclaved
separately, and 200 mg of ampicillin was added per liter
through a syringe filter (0.2 m). The feed media used
for fed-batch studies contained, per liter, (a) 1.5 g of
(NH4)2SO4, yeast extract (variable amounts); (b) 1.5 g of
MgSO47H2O; and (c) glucose (variable amounts). The
feed media contained 1 g of ampicillin/L, and components
a, b, and c were also autoclaved separately.
Bioreactor Operation. All batch and fed-batch
bioreactor experiments were conducted in 5-L Bioflo III
laboratory fermenters (New Brunswick Scientific, U.S.).
Unless otherwise mentioned, the bioreactors were oper-
ated at pH 6.75 and 37 C. Dissolved oxygen was
controlled above 40% air saturation to avoid oxygen
limitation. Data acquisition and control of various
operating parameters (e.g., temperature, dissolved oxy-
gen concentration, pH, rpm, and the medium flow rate)
were effectively achieved by using the Bioflo IIIs elec-
tronic control module and the software, AFS (Advanced
Fermentation System).
Single-Stage Fed-Batch Process. When the specific
growth rate (or glucose uptake rate) of E. coli exceeds a
certain critical rate, acetic acid formation increases and
it becomes a significant inhibitory product. The specific
growth rate of E. coli therefore needs to be carefully
controlled to achieve a high cell density in fed-batch
cultures. In this study, the fed-batch process was oper-
ated with the feedback control of media feed by pH-stat
method. In the pH-stat control, the media feed is
controlled by on-off mode depending on the culture pH,
i.e. the media feed occurs while the culture pH is above
a certain value, pHset (e.g., 6.75 in this study) and the
media feed is switched off when the culture pH goes down
Biotechnol. Prog., 1997, Vol. 13, No. 3
below the pHset. The volumetric media feed rate during
the feeding-on period was stepwise increased as the cell
concentration increases. As soon as the recombinant cell
concentration reached an appropriate concentration (70-
80 g/L), the recombinant gene expression was started by
feeding the postinduction media to the cell broth. The
T7 promoter inducer, IPTG was injected separately to
the culture (variable amounts) through a syringe filter
at the beginning of the postinduction feed. Since sub-
stantial cell growth has not been observed while the
recombinant gene is expressed, the volumetric feed rate
of the postinduction media was maintained at a constant
rate.
Two-Stage, Cyclic Fed-Batch Process. The two-
stage, cyclic fed-batch process is a modified fed-batch
process that a portion of the cell broth is transferred to
an induction tank while leaving some cell broth in the
growth tank. The volume of liquid in the growth tank is
replenished to its pre-transfer volume while product
formation is taking place in the induction tank. The
mathematical analysis of the cyclic fed-batch process was
previously published by Keller and Dunn (1978) and Mori
et al. (1983). The key fermentation variables to operate
the fed-batch reactor by the two-stage cyclic mode are
defined as (1) media feed rate (F1) for growing the
recombinant cells at constant cell concentration and
specific growth rate in growth tank; (2) glucose con-
centration (SF) in feed media in growth tank; (3) volume
of the cell broth (V0i) transferred to induction tank; and
(4) volume of the cell broth (V0g) remaining in growth
tank after volume transfer. The above key variables were
mathematically defined by the following design equa-
tions:
dV
F1 ) ) V0g exp(t) (1)
dt
where is the specific growth rate of the recombinant
cells and t is the cultivation time in the growth tank after
the volume transfer;
Xg
SF ) (2)
Yx/s
where Xg is the recombinant cell concentration main-
tained constant in growth tank and Yx/s is the biomass
yield on glucose;
V0i ) V0g[exp(tind) - 1] (3)
where tind is the time required for induction; and
Vg,max
V0g ) (4)
exp(tind)
where Vg,max is the maximum working volume of the
growth tank. Using the design equations described above
(eqs 1-4), the specific growth rate and the cell concentra-
tion of the recombinant E. coli were kept constant in the
growth tank by controlling volumetric feed rate and
substrate concentration of the feed to the growth tank
and transfer volume of growth culture to induction tank.
Conditions in the induction tank, such as pH, volumetric
feed rate, and feed medium composition, were operated
independently from conditions in the growth tank.
Analytical Methods. The glucose concentration was
measured by a YSI glucose analyzer (model 2300 STAT
PLUS, Yellow Springs Instruments, Yellow Springs, OH),
and the acetic acid concentration was measured by
employing a acetate assay reagent (Boehringer Mann-
251
heim) as per the procedure suggested by the supplier.
The optical density of each culture sample was measured
at 600 nm using a Pharmacia Ultrospec III spectropho-
tometer. The recombinant cell mass concentration was
then obtained using a previously developed correlation
between optical density and dry cell mass concentration.
Expression levels of human mini-proinsulin were meas-
ured by subjecting samples taken from the fermenter to
denaturing gradient sodium dodecyl sulfate-polyacryl-
amide gel electrophoresis (SDS-PAGE) (10-20% tricine
gel, NOVEX, San Diego, CA). Bovine serum albumin
(BSA) (3 g) was separately loaded on each gel with
culture samples as standard for protein quantification.
The resulting protein bands were scanned with an
Ultroscan XL (Pharmacia LKB Biotechnology, Sweden)
laser densitometer. From this analysis, the percentage
of recombinant protein from the total cellular protein was
determined and the amount of human mini-proinsulin
produced was also calculated from a previously developed
correlation between BSA concentration and the measured
band area. The plasmid copy number was determined
by separating the whole DNA lysate from a specific
quantity of cell suspension via 0.8% agarose gel electro-
phoresis and then by scanning the photonegatives of
stained gels with the laser densitometer mentioned above
(Nugent et al., 1983).
Results and Discussion
High Cell Density Culture of Recombinant E. coli
BL21(DE3)[pET-3aT2M2]. Since it has been observed
that the cloned gene product, mini-proinsulin signifi-
cantly inhibits the host cell growth due presumably to
its toxicity (to be discussed later), it is essential to control
the time and extent of plasmid gene expression using the
inducible plasmid promoter, T7. That is, the cells are
grown first to a high concentration in the absence of
recombinant gene expression (preinduction, or growth
phase) and then the T7 promoter system is induced to
synthesize the mini-proinsulin (postinduction, or product
formation phase). For achieving the high cell density
cultures of the recombinant E. coli, as illustrated earlier,
the growth phase of fed-batch process was operated with
controlling the medium feed by a pH-stat method. As
the composition of preinduction feed (nutrient glucose (G)
and yeast extract (YE) concentration) is an important
factor affecting the recombinant cell growth, the biomass
production was investigated at various compositions in
fed-batch cultures (data not shown) and an optimal G
and YE concentration leading to the maximum biomass
production in the growth phase was determined as 250
g of glucose/L and 204 g of yeast extract/L. Consequently,
via the pH-stat control of the optimized medium feed, it
was possible to attain the high cell density culture of
recombinant E. coli where the biomass concentration
increased to more than 70 g/L while acetate concentration
was maintained below 1 g/L during the whole growth
phase.
Optimum Induction Strategy. The volumetric yield
of a recombinant product depends on both the biomass
concentration and the specific cellular product yield.
Whereas fed-batch processes primarily focus on increas-
ing the biomass concentration, the cultivation conditions
can also affect the specific cellular product yield. De-
pending on the characteristics of particular host-vector
system, some cultivation conditions are closely related
to the regulation of cloned gene expression and, therefore,
can have an important influence on the product yield of
fed-batch processes employing recombinant microorgan-
isms. The optimum induction strategy can then be
252 Biotechnol. Prog., 1997, Vol. 13, No. 3

Table 1. Regulation of Specific Cell Growth Rate by the Exponential Feeding of Preinduction Media in a Cyclic Growth
Experiment
dry cell mass, g/L specific
operation mode time, h (culture volume, L) growth rate, h-1
batch 0.00 0.40
2.00 0.96 0.441
4.00 5.09 0.835
6.00 10.85 0.379
6.50 11.79 0.165
fed-batch (pH-stat) 10.00 20.95 0.192
13.00 31.80 0.173
16.00 43.46 0.151
18.75 52.15 0.126
25.50 73.61 0.140
harvest cell culture
fed-batch (exponential feeding: a ) 0.12 h-1) 25.50 73.61 (2.00) 0.140
26.50 74.62 (2.25) 0.134
28.00 75.30 (2.70) 0.126
30.00 74.62 (3.43) 0.116
32.50 73.95 (4.63) 0.116
harvest cell culture
fed-batch (exponential feeding: a ) 0.10 h-1) 32.50 73.95 (2.00) 0.116
34.50 73.27 (2.44) 0.095
37.00 72.25 (3.14) 0.094
39.50 71.23 (4.03) 0.094
41.00 71.57 (4.68) 0.103
harvest cell culture
fed-batch (exponential feeding: a ) 0.08 h-1) 41.00 71.57 (2.00) 0.103
43.50 72.93 (2.44) 0.088
46.00 72.59 (2.98) 0.078
49.00 71.91 (3.79) 0.077
51.00 71.23 (4.45) 0.075
51.50 71.23 (4.63) 0.080
a Specific growth rate used in eq 1 which determines the volumetric feed rate of growth media.

determined by taking the effects of various cultivation
conditions on the recombinant gene expression into
account.
Preinduction Specific Growth Rate. From the
previously published reports, several different correla-
tions between specific product formation and specific
growth rate have been observed in the recombinant E.
coli cultures. Siegel and Ryu (1985) and Curless et al.
(1990) observed an increase in the recombinant product
(trpA1 protein and R consensus interferon (IFN-RCon1),
respectively) yields with an increase in the preinduction
specific growth rate. Jung et al. (1988) reported a lower
cell concentration and specific yield of interleukin-1 in
constantly fed cultures compared to exponentially fed
cultures. The average specific growth rates of the
constant feed and exponential feed culture were 0.14 and
0.36 h-1, respectively. This may also be due to a
correlation between the lower specific growth rate and
specific recombinant product yields in the constant feed
culture. However, Seo and Bailey (1985) demonstrated
a decline in the concentration of recombinant protein
product, -lactamase, when the growth rate was in-
creased by a change in media composition in batch
cultures of a plasmid-harboring E. coli HB101. Riesen-
berg et al. (1990) reported that the cellular level of human
interferon R1 (IFNR1) was higher at lower growth rates
in glucose-limited chemostat studies. On the other hand,
several groups (Bech Jensen and Carlsen, 1990; Zabriskie
et al., 1987) have observed no correlation between the
specific growth rate and specific recombinant product
formation. Seemingly, the relationship depends on the
characteristics of particular recombinant system, e.g.
promoter system, host-vector interaction, toxicity of the
recombinant product, etc. The dependency of the recom-
binant product formation on the preinduction specific
growth rate is very important because it directly relates
to the choice of feeding control method of preinduction
media. That is, if the preinduction specific growth rate
critically influences the specific/volumetric yield of re-
combinant product, the growth rate of the recombinant
cells should be precisely regulated during the growth
phase, which is only possible by the exponential feeding
method, not by the feedback control method such as the
pH-stat. In present study, the effect of the preinduction
specific growth rate on recombinant product formation
was investigated in fed-batch cultures. The pH-stat
method was first employed to achieve a high cell density
culture and when the cell concentration reached 70-75
g/L, feeding was switched to exponential to regulate the
recombinant cell growth at a desired growth rate at a
constant cell concentration. The volumetric feed rate and
the substrate feed concentration used for the exponential
feeding were determined by the design equations, eqs
1-4. As shown in the result of a cyclic growth experi-
ment (Table 1), the specific growth rate was successfully
regulated at various desired rates at a constant cell
concentration. The tested range of specific growth rate
was below 0.14 h-1 because the regulation of the specific
growth rate failed above 0.14 h-1 due to the increase in
the acetic acid production during the exponential feeding
period (data not shown). The IPTG concentration in the
culture was 1.0 mM at the time of induction and the
optimized preinduction feed media was used as post-
induction feed at this time. The feed rate of postinduc-
tion media was determined at 0.55 mL/h per unit cell
mass (g) of recombinant E. coli at the time of induction
and was maintained constant afterwards. As shown in
Figure 2, no distinct correlation between the specific
growth rate and recombinant product formation was
observed. This implies that the feeding strategy of
preinduction media is not a critical factor in regulating
the recombinant gene expression and thus the pH-stat
is an applicable method of preinduction feed control for
growing the recombinant cells prior to starting the mini-
proinsulin synthesis at high cell density cultures.
Biotechnol. Prog., 1997, Vol. 13, No. 3
Figure 2. Relationship between preinduction specific growth
rate and mini-proinsulin production in fed-batch cultures.
Induction Temperature. The culture temperature
in the product formation phase is an important cultiva-
tion condition which should be considered in optimizing
the recombinant gene expression. Primbrose et al. (1984)
reported that temperature affects the plasmid stability
in the recombinant E. coli cultures. Kennell (1986)
demonstrated that the rate of mRNA degradation is first
order and decreases with temperature. Thus, it is
possible that lowering the culture temperature may be
a simple and potentially important method for increasing
protein production. The temperature effect on the re-
combinant product formation was investigated in the
present study: the culture temperature of growth phase,
37 C, was shifted to 34 and 30 C at the time of induction
in each experiment. The results show that lowering the
induction temperature is unfavorable for the mini-
proinsulin production and that the cell growth in post-
induction phase is inversely proportional to the amount
of recombinant gene expression, which is probably due
to the toxicity of mini-proinsulin (Figure 3). The severe
inhibition of recombinant cell growth in the postinduction
phase at 30 C does not seem to result entirely from the
suboptimal growth temperature, and hence, it is pre-
sumed that the very low amount of mini-proinsulin is
toxic to the host cell. It is interesting to find that the
mini-proinsulin synthesis was almost completely re-
pressed at 30 C and the employed T7-based expression
system seems very temperature-sensitive. With this
promoter system, cultivating the recombinant cells at
temperature below growth optimum may be potentially
useful for tighter regulation of gene expression in produc-
tion of highly toxic recombinant proteins.
Nutrient Composition of Postinduction Media.
The mini-proinsulin yield in the induced cell culture is
significantly affected by the nutrient (G and YE) com-
position of the postinduction feed. A statistical method
for optimizing medium composition, response surface
methodology (RSM), was employed to determine the
optimal G and YE concentration maximizing the mini-
proinsulin production. The RSM based on factorial
experiments is a statistical approach to study the effect
of test variables on measured responses: the G and YE
concentrations in the postinduction feed (test variables)
were systematically varied and the expression level of
mini-proinsulin (response) was analyzed by SDS-PAGE
253
Figure 3. Effect of culture temperature on recombinant mini-
proinsulin production. b, 2, 9: dry cell mass (g/L); O, 4, 0:
product concentration (g of mini-proinsulin/L).
at each composition (Table 2). From the response surface
presented in Figure 4, it is evident that there exists a
unique maximum on the convex surface, and from
canonical analysis of the response surface, the maximum
value of expression level is reached at a (G,YE) point,
i.e. (274.45,210.88). Consequently, the optimal glucose
and yeast extract concentrations in the postinduction feed
were determined at 274 and 211 g/L, respectively.
Inducer (IPTG) Amount. The host BL21 (DE3)
contains a single copy of the T7 RNA polymerase gene
in chromosome under control of the inducible lacUV5
promoter, and therefore, the addition of IPTG induces
the lacUV5 promoter to produce T7 RNA polymerase,
which in turn initiates high-level expression of the target
gene in the plasmid. It is now generally accepted that
in the inducible promoter system the transcription is
initiated when a specific small molecule (inducer) binds
to the repressor protein which is bound to the operator
(a DNA sequence overlapping the promoter) in the
absence of the inducer and inhibits the interaction of
RNA polymerase with the promoter. The amount of
inducer required to titrate the repressor molecules is
proportional to the total cell mass and the optimal specific
concentration of the inducer therefore needs to be deter-
mined for maximizing the mini-proinsulin synthesis at
any cell concentration. In a range of specific amount of
inducer (IPTG) added to the recombinant cell broth (3.26
10-3 to 5.11 10-2 mmol/g of cell), the recombinant
gene expression was analyzed in the fed-batch cultures
with the feed of the optimized postinduction media
(Figure 5). The highest yield of mini-proinsulin was
achieved at a specific IPTG injection amount, i.e. 3.03
10-2 mmol/g of cell.
Feed Rate of Postinduction Media. Since it has
been observed that the cell division is nearly stopped
after the recombinant protein synthesis is started, it
should be appropriate to feed the postinduction media
at a constant rate. If this statement is generalized for
any cell density (x) and volume (V) being induced with a
given feed media, F ) aXV, where F is the volumetric
feed rate of postinduction media and a is a proportionality
constant relating feed rate to total biomass, defined as
specific feed rate of postinduction media (mL/(h/g of cell))
in this paper. The volumetric feed rate of postinduction
media is directly related to the availability of nutrient
glucose or amino acids in cell broth. Adams and Hatfield
(1984) have shown that starvation of essential amino
acids causes an increase in the plasmid copy number. The
increase in the plasmid copy number enhances the level
of mRNA, and thereby, the formation of a protein product
254 Biotechnol. Prog., 1997, Vol. 13, No. 3

Figure 5. Effect of specific injection amount of IPTG on mini-

proinsulin yield.
variation of plasmid copy number in recombinant cell.
On the other hand, Goldberg and St. John (1976) and
Tsai et al. (1987) have observed that starvation of the
culture for carbon or amino acids results in increased
proteolysis and Talmadge and Gilbert (1982) have re-
ported that the half-life of the preproinsulin in the E.
coli cytoplasm is less than 2 min. Since however the
mini-proinsulin is finally aggregated to form the insoluble
inclusion bodies, the proteolytic degradation of this small
protein can be prevented although not completely avoided
until the formation of the inclusion bodies is finished in
Figure 4. Relationship between the expression level of mini- the cytoplasm. Fed-batch experiments were conducted
proinsulin (% of total proteins) and the nutrient composition, to determine the optimal value of the proportionality
represented as a response surface using multiple regression constant, a, yielding the highest recombinant gene
technique using the following model: Y ) b0 + b1G + b2YE + expression. A range of constant a (i.e., specific feed rate
b3G2 + b4YE2 + b5GYE where Y is the predicted value of
expression level of mini-proinsulin; G and YE refer to coded of postinduction media) was examined with regard to the
amounts or actual concentrations (g/L) of nutrient glucose and influence on the mini-proinsulin production using the
yeast extract, respectively; and b values are the computed effective amount of IPTG and optimized postinduction
regression coefficients, representing effects of nutrient glucose media. Evidently, at 0.45 mL/(h/g of cell), the mini-
and yeast extract. Calculation results of regression coefficients proinsulin concentration was maximized in the fermen-
are b0 ) 15.0306, b1 ) 0.1044, b2 ) -1.0994, b3 ) -6.1059, b4 tation culture (i.e., 7 g of mini-proinsulin/L) (Figure 6)
) -2.9391, b5 ) 1.0294, R2 ) 0.9995. (a) Three-dimensional
surface. (b) Contours of expression level (numbers in the figure and the highest volumetric productivity (i.e., 0.175 g of
indicate the expression level of mini-proinsulin). mini-proinsulin/(L/h)) was also obtained at this condition.
Immediately after the postinduction phase begins, the
Table 2. Variation in Coded Value and Concentration of plasmid copy number increases many fold compared to
G and YE for Optimizing Postinduction Feed that of preinduction phase (Figure 7a). Due presumably
Composition by RSM, and Mini-Proinsulin Production at to the toxicity of mini-proinsulin to the host cell, even
Each Composition
an initial low level of synthesized mini-proinsulin may
coded nutrient be enough to hold the cellular division while the other
values concentration (g/L) mini-proinsulin production cellular machineries such as DNA replication and RNA/
G YE G YE (% of total proteins) protein synthesis are still functioning though less active.
1 1 300 250 13.7 As expected earlier in this section, the variation pattern
1 -1 300 150 12.2 of the plasmid copy number apparently depends on the
-1 1 200 250 7.7 specific feed rate of the postinduction media (i.e., constant
-1 -1 200 150 8.0 a), which may result from difference in the availability
0 0 250 200 14.6 of key nutrients in the culture both. Figures 6 and 7a
2 0 350 200 8.8
0 150 200 13.3
demonstrate that the maximum plasmid copy number at
each feed rate of postinduction media is approximately
-2
0 2 250 300 10.0
0 -2 250 100 4.9 at the same level (140-165 copies per chromosome), and
that as earlier the plasmid copy number reaches maxi-
could be increased. It is reasonable to presume that the mum in the postinduction phase, the higher expression
degree of the nutrient (mainly, glucose and amino acids) level of the recombinant protein is achieved. This is
starvation, which is determined by the volumetric feed probably because the viability of induced cells is decreas-
rate of postinduction media, will significantly affect the ing as the recombinant protein synthesis goes on and
Biotechnol. Prog., 1997, Vol. 13, No. 3
Figure 6. Effect of specific feed rate of postinduction media
(i.e., constant a in equation F ) aXV) on mini-proinsulin
production: b, yield (g of mini-proinsulin/g of cell); 9, product
concentration (g of mini-proinsulin/L).
higher gene expression can be achieved with more viable
cells. From Figure 7a,b, at an optimal specific feed rate
of postinduction media (i.e., 0.45 mL/(h/g of cell)), the
plasmid copy number starts to decrease after it reaches
maximum and the maximum expression level of the
recombinant mini-proinsulin is achieved while the plas-
mid copy number is decreasing (i.e., several hours after
the maximum plasmid copy number is obtained). It seems
likely that the mRNA produced by T7 RNA polymerase
can rapidly saturate the post-transcriptional machinery
of E. coli, so that the rate of protein synthesis from such
an mRNA depends primarily on the post-transcription
efficiency.
Production of Human Mini-Proinsulin via Two-
Stage, Cyclic Fed-Batch Process. As previously il-
lustrated in this paper, the recombinant product synthe-
sis does not depend on the preinduction specific growth
rate in an investigated range of cell growth rate (0.05-
0.14 h-1). This is an advantageous feature in that the
cell growth rate and hence the growth-cycle period in the
growth tank can be adjusted to the demand of total
operating time required in the induction tank. For
example, to ensure the sufficient operating time in
induction tank, the V0g in eq 4 can be substituted by a
minimum working volume of the fermenter and then a
specific cell growth rate to be regulated in the growth
tank is determined from the same equation. In the
present study, the culture volume of the growth tank
changes from minimum to maximum working volume
every growth cycle and a specific cell growth rate in the
growth tank was determined so that about 14 h of total
operating time (tind) can be allowed in the induction tank.
The other operating variables used for two-stage, cyclic
fed-batch operation were determined by eqs 1-3.
As shown in Figure 8, cycling was executed five times.
The recombinant cell concentration was maintained
between 72 and 80 g/L in the growth tank, and acetate
was not accumulated in the fed-batch culture for each of
five cycles (Figure 8a). Figure 8b represents that the
postinduction dry cell mass ranged between 57 and 73
g/L and that, at every induction cycle of two-stage, cyclic
operation, higher (10-44%) volumetric productivity of
255
Figure 7. (a) Variations in plasmid copy number during
postinduction phase at various specific feed rates of postinduc-
tion media: 9, 0.45 mL/(h/g of cell); O, 0.35 mL/(h/g of cell); 4,
0.65 mL/(h/g of cell). (b) Profiles of mini-proinsulin production
(O) and plasmid copy number (9) at an optimal specific feed
rate (0.45 mL/(h/g of cell)) of postinduction media.
mini-proinsulin was obtained than the maximum volu-
metric productivity (0.175 g/(L/h)) achieved in single-
stage fed-batch process. It is therefore evident that the
two-stage, cyclic fed-batch process is preferable to the
single-stage process in increasing the volumetric produc-
tivity of mini-proinsulin and will contribute significantly
toward increasing overall production rate of human
insulin. For increasing further the productivity of the
two-stage, cyclic fed-batch processes, the composition and
volumetric feed rate of postinduction media could be
optimized again. As described in Figure 8, the prein-
duction feed composition in the two-stage, cyclic fed-batch
process is different from that used in the single-stage
process whereas the postinduction feed medium used in
both processes is the same. The postinduction feed
composition was previously optimized for the recombi-
nant cell cultures grown with the preinduction feed, the
composition of which was optimized for maximizing the
biomass production in the growth phase of single-stage
process. The optimum composition of postinduction feed
may depend on the nutritional conditions of pre-
induction culture, and therefore, the postinduction feed
composition could be reoptimized for increasing the
protein yield and hence productivity at the two-stage,
cyclic fed-batch process. Since, as demonstrated earlier,
the changes in volumetric feed rate of postinduction
media lead to the variations in plasmid copy number
which in turn significantly influence the mini-proinsulin
256
Figure 8. (a) Profiles of dry cell mass and culture volume in
growth tank at operating a two-stage, cyclic fed-batch process:
b, time of culture transfer. (b) Preinduction and postinduction
dry cell mass and volumetric productivity of recombinant mini-
proinsulin versus cycle number in induction tank: b, preinduc-
tion dry cell mass (g/L) at the time of culture transfer; 9,
preinduction dry cell mass (g/L) when cell growth rate is
restored in induction tank (the additional growth period (2-3
h) was given to restore the cell growth rate up to around pre-
transfer level before the cloned gene expression after the growth
culture transfer to induction tank.); 2, postinduction dry cell
mass (g/L); bars, volumetric productivity of mini-proinsulin (g
of mini-proinsulin/(L/h)) at each induction cycle; - - -, maximum
volumetric productivity of mini-proinsulin obtained in single-
stage fed-batch process (0.175 g of mini-proinsulin/(L/h)). The
values of operating and kinetic variables used for two-stage,
cyclic fed-batch operation are ) 0.078 h-1, V0g ) 1.5 L, Vg,max
) 4.5 L, Xg ) 75 g/L, YX/S ) 0.42, SF ) 179 g/L, and tind ) 14 h.
synthesis, the feed rate of reoptimized postinduction
media could be also optimized again.
The two-stage, cyclic process requires more feed media
tanks because of a continuous need for feed than a single-
stage fed-batch process, but much smaller fermenters.
Fulfilling the cycling procedure was difficult because
manipulations were carried out manually in the labora-
tory. However, in industrial fermentation plants, the
process could be automated by using a microprocessor
to control both feed pumps, the volume transfer pump,
the volume transfer valve, and the drain valve on the
induction tank.
Conclusions
For the purpose of enhancing the human insulin
production, mini-proinsulin, a mutated protein of natural
human proinsulin, was stably expressed in a small fusion
protein in the fed-batch cultures at high cell density of
Biotechnol. Prog., 1997, Vol. 13, No. 3
recombinant E. coli BL21(DE3)[pET-3aT2M2]. The syn-
thesis of mini-proinsulin seems inhibitory to the host cell
growth, and the T7 promoter system employed in this
study was shown useful in controlling the time and extent
of plasmid gene expression.
The correlation between preinduction specific growth
rate and recombinant gene expression is important
because it is directly related to the feeding strategy of
preinduction media. Since no general correlation was
found, either the pH-stat or exponential feeding method
can be used for preinduction feed control in fed-batch
cultures, and furthermore, it becomes possible in two-
stage fed-batch processes to adjust a growth-cycle period
in the growth tank to the demand of total operating time
required for maximum protein yield in the induction
tank. Low induction temperature was apparently un-
favorable for mini-proinsulin synthesis, and the T7-based
pET expression system was almost completely repressed
at 30 C. With this promoter system, the low cultivation
temperature may be potentially effective for tighter
regulation of cloned gene expression in case of highly
toxic recombinant protein synthesis. The relationship
between nutrient (glucose and yeast extract) compositions
of postinduction media and expression level of mini-
proinsulin was expressed by a response surface, the
analysis of which clearly showed that there exists a
unique optimal composition maximizing the recombinant
protein expression level. The changes in volumetric feed
rate of postinduction media caused the variations in
plasmid copy number profile and in turn critically
influenced the recombinant gene expression. It is con-
cluded that the mini-proinsulin yield depends on how
soon the copy number reaches maximum after the gene
expression is initiated because the viability of induced
cells diminishes as the recombinant protein synthesis
proceeds and higher-level expression can be achieved
with more viable cells. It seems likely that the rate of
mini-proinsulin synthesis depends primarily on the post-
transcription efficiency. The two-stage, cyclic process was
feasible for producing recombinant mini-proinsulin from
E. coli for an extended period of time at a high cell
density. Compared to a continuous fermentation process,
the two-stage fed-batch process produces homogeneous
cell populations with respect to induction time. With
respect to volumetric reactor productivity, the two-stage,
cyclic fed-batch process was apparently more productive
than the single-stage fed-batch fermentation process.
Consequently, the mini-proinsulin yield and productiv-
ity were significantly enhanced by optimizing the key
cultivation variables and then by applying the efficient
strategy of bioreactor operation, i.e. two stage, cyclic
operation in the high cell density fed-batch cultures. The
nutrient composition and volumetric feed rate of post-
induction feed medium may need to be optimized again
for increasing further the productivity of two-stage, cyclic
fed-batch process, which will be the next stage to be
explored in the future.
Literature Cited
Adams, C. W.; Hatfield, G. W. Effects of promoter strengths and
growth conditions on copy number of transcription-fusion
vectors. J. Biol. Chem. 1984, 259, 7399-7403.
Bech Jensen, E.; Carlsen, S. Production of recombinant human
growth hormone in Escherichia coli: Expression of different
precursors and physiological effects of glucose, acetate, and
salts. Biotechnol. Bioeng. 1990, 36, 1-11.
Chen, J.-Q.; Zhang, H.-T.; Hu, M.-H.; Tang, J.-G. Production of
human insulin in an E. coli system with met-lys-human
proinsulin as the expressed precursor. Appl. Biochem. Bio-
technol. 1995, 55, 5-15.
Biotechnol. Prog., 1997, Vol. 13, No. 3
Curless, C.; Pope, J.; Tsai, L. Effect of preinduction specific
growth rate on recombinant alpha consensus interferon
synthesis in Escherichia coli. Biotechnol. Prog. 1990, 6, 149-
152.
Georgiou, G. Optimizing the production of recombinant proteins
in microorganisms. AIChE J. 1988, 34 (8), 1233-1248.
Goldberg, A. L.; St. John, A. L. Intracellular protein degradation
in mammalian and bacterial cells: part 2. Ann. Rev. Biochem.
1976, 45, 747-803.
Jung, G. D.; Denefle, P.; Becquart, J.; Mayaux, J.-F. High-cell
density fermentation studies of recombinant Escherichia coli
strains expressing human interleukin-1. Ann. Inst. Pasteur/
Microbiol. 1988, 139, 129-146.
Keller, R.; Dunn, I. J. Computer simulation of the biomass
production rate of cyclic fed-batch continuous culture. J. Appl.
Chem. Biotechnol. 1978, 28, 784-790.
Kennell, D. E. The instability of messenger RNA in bacteria.
In Maximizing Gene Expression; Reznikoff, W., Gold, L., Eds.;
Butterworths: Boston, MA, 1986.
Kroeff, E. P.; Owens, R. A.; Campbell, E. L.; Johnson, R. D.;
Marks, H. I. Production scale purification of biosynthetic
human insulin by reversed-phase high-performance liquid
chromatography. J. Chromatogr. 1989, 461, 45-61.
Lee, J.; Parulekar, S. J. Enhanced production of R-amylase in
fed-batch cultures of Bacillus subtilis TN106[pAT5]. Biotech-
nol. Bioeng. 1993, 42, 1142-1150.
Lee, J.; Parulekar, S. J. Periodic operation of continuous
recombinant cultures improves antibiotic selection. Chem.
Eng. Sci. 1996, 51 (2), 217-231.
Marston, F. A. O. The purification of eukaryotic polypeptides
synthesized in Escherichia coli. Biochem. J. 1986, 240, 1-12.
Mori, H.; Yamane, T.; Kobayashi, T.; Shimizu, S. Comparison
of cell productivities among fed-batch, repeated fed-batch and
continuous cultures at high cell concentration. J. Ferment.
Technol. 1983, 61 (4), 391-401.
Nugent, M. E.; Primrose, S. B.; Tacon, W. C. Chapter 21. The
stability of recombinant DNA. In Developments in Industrial
Microbiology (Proceedings of the Thirty-Ninth General Meet-
ing of the Society for Industrial Microbiology, Held at St. Paul,
Minnesota August 14-20, 1982); SIM: Arlington, VA, 1983;
Vol. 24, pp 271-285.
Parulekar, S. J.; Lee, J. Structure analysis of continuous
cultures subject to periodic medium tuning. Chem. Eng. Sci.
1993, 48 (17), 3007-3035.
Petrides, D.; Sapidou, E.; Calandranis, J. Computer-aided
process analysis and economic evaluation for biosynthetic
human insulin productionsa case study. Biotechnol. Bioeng.
1995, 48, 529-541.
257
Primbrose, S. B.; Derbyshire, P.; Jones, I. M.; Robinson, A.;
Ellwood, D. C. The application of continuous culture to the
study of plasmid stability. Contin. Cult. Biotechnol. Med.
Environ. 1984, 8, 213.
Raskin, P.; Clements, R. S. The use of human insulin derived
from Bakers yeast by recombinant DNA technology. Clin.
Ther. 1991, 13 (5), 569-578.
Riesenberg, D.; Menzel, K.; Schulz, V.; Schumann, K.; Veith,
G.; Zuber, G.; Knorre, W. A. High cell density fermentation
of recombinant Escherichia coli expressing human interferon
alpha 1. Appl. Microbiol. Biotechnol. 1990, 34, 77-82.
Seo, J.-H.; Bailey, J. E. Effects of recombinant plasmid content
on growth properties and cloned gene product formation in
Escherichia coli. Biotechnol. Bioeng. 1985, 27, 1668-1674.
Shin, H. C.; Chang, S. K.; Kim, D. Y.; Kang, S. J. Use of a beta-
sheet forming amino acid leader sequence for the production
of proteins and peptides. Korean Patent Application No. 94-
7018, 1994.
Shin, H. C.; Chang, S. K.; Kim, D. Y. A process for preparing
human proinsulin by recombinant DNA technology. Korean
Patent Application No. 95-2751, 1995.
Siegel, R.; Ryu, D. D. Y. Kinetic study of instability of recom-
binant plasmid pPLc23trpA1 in E. coli using two stage
continuous culture system. Biotechnol. Bioeng. 1985, 27, 28-
33.
Talmadge, K.; Gilbert, W. Cellular location affects protein
stability in Escherichia coli. Proc. Natl. Acad. Sci. U.S.A.
1982, 81, 1830-1833.
Tsai, L. B.; Mann, M.; Morris, F.; Rotger, C.; Fenton, D. The
effect of organic nitrogen and glucose on the production of
recombinant human insulin-like growth factor in high cell
density Escherichia coli fermentations. J. Ind. Microbiol.
1987, 2, 181-187.
Williams, D. C.; Van Frank, R. M.; Muth, W. L.; and Burnett,
J. P. Science 1982, 215, 687-689.
Yee, L.; Blanch, H. W. Recombinant protein expression in high
cell density fed-batch cultures of Escherichia coli. Bio/
Technology 1992, December 10, 1550-1556.
Zabriskie, D. W.; Arcuri, E. J. Factors influencing productivity
of fermentations employing recombinant microorganisms.
Enzyme Microb. Technol. 1986, December 8, 706-717.
Accepted March 7, 1997.X
BP970018M
X Abstract published in Advance ACS Abstracts, April 15, 1997.