Opinion

TRENDS in Biotechnology Vol.20 No.3 March 2002

Optimizing scale-up
fermentation
processes
Michel Thiry and Doriano Cingolani
There are many aims associated with the optimization of fermentation
processes. Optimization is expected to increase the yield of the final product
but the process must be compliant with good manufacturing practices, the
available equipment and the expected final scale of operation. Dealing with
genetically modified microorganisms that overproduce recombinant protein
has the advantage that the vast majority of the processes use only three
different species, namely Escherichia coli, Saccharomyces cerevisiae and
Pichia pastoris. Standard processes for each organism are described in
textbooks and serve as a basis for the development of a tailored process.
This article outlines the general philosophy that we have devised to ensure an
efficient approach of scaling up fermentation processes for biopharmaceutical
purposes, in a multidisciplinary environment.

Michel Thiry*
Doriano Cingolani
Eurogentec S.A.,
Parc scientifique du Sart
Tilman, B-4102 Seraing,
Belgium
*e-mail:
m.thiry@eurogentec.com

The optimization of the fermentation process is

included in the strategic analysis of the viability of a
project. Optimization takes place once the feasibility
of the production in the selected organism has been
demonstrated. This implies that an expression
system has been constructed and that, at least from
a theoretical viewpoint, the system should be
regarded as the optimum. Before starting long and
expensive optimization work, it is important that
the stability of the strain is established, at least for
the number of generations necessary for cell banking
and largest-scale fermentation, including the
pre-cultures. Expression systems based on plasmids
are sometimes unstable several parameters affect
the segregational stability of plasmids. The stability
is different for each plasmid and depends on the host
strain [1]; high instabilities are correlated with low
copy number plasmids [2]. The size of the inserted
DNA affects the stability; the larger the plasmid is,
the less stable it is [3]. Culture conditions, such as
temperature [4], the composition of the media and
the growth rate [5], can also modify the plasmid
stability. Plasmid loss is significant over the postexponential growth phase [6] and plasmid instability
is enhanced during the period of gene expression.
Antibiotics are often used to stabilize plasmids.
The compensation of an auxotrophic defect of the
host coded on the plasmid [7] is more suitable than
using antibiotics for regulatory reasons. However,
the preparation of media for auxotrophic strains is
laborious. The plasmid can be stabilized by the
insertion of the parB (hok/sok) locus, which kills the
plasmid-free cells by segregation [8]. An alternative
is the integration of the insert in the hosts
http://tibtech.trends.com

103

chromosome. This is often the case for Pichia pastoris

constructions but gene integration can reduce the
expression level [9].
The choice of the producing strain and the
expression vector will depend on the constitutive or
inducible expression into the cytoplasmic
compartment or the secretion into the outer medium.
If the strain has been acquired outside the laboratory,
the quality of the documentation of the source of the
original strain and the cloning procedure must be
evaluated and approved by a competent quality
assurance service.
The codon usage of the gene of interest should be
optimized to facilitate the expression in the chosen
microorganism. A screening of expression in a
shake-flask must be performed immediately to find
a clone that produces the recombinant protein at a
high level of expression.
Optimization of culture conditions

Given that the main aim of optimization is to

maximize the production, this process can be initiated
only once a laboratory-scale purification process and
a minimum set of quality control tools are available
to quantify and assess the quality of the product.
The fermentation protocol affects the impurity profile
and thus impacts strongly on the efficacy of the
downstream processing (DSP). Similarly, the
conditions of fermentation can determine whether
the protein of interest will be in its soluble or
insoluble form, which will deeply affect the DSP and
consequently the quality and yield of the purified
product. Overproduction at high yield can also lead to
the depletion of certain factors that are necessary for
good conformation of the protein. The fermentation
must thus be developed concomitantly with the
purification because they will affect each other.
The quality of the communication between the
developers of the fermentation and the DSP is pivotal
to the success of the scaling up the process (Fig. 1).
Optimization of the fermentation process can be
conducted either by changing one factor at a time or
by varying several factors at the same time and
looking for interactions using statistical analysis.
The statistical design of experiments is an organized
approach that gives more reliable information per
experiment than unplanned approaches. Statistical
data analysis enables the visualization of the
interactions among experimental variables, leading
to predictions of the response data in areas not
directly covered by experimentation [10]. One of the
most important operations for the investigator is to
define the variables that could affect the productivity
or the quality of the product.
Fermentation can be performed in batch, fed-batch
or continuous mode. Continuous culture is not
common in the pharmaceutical industry because the
probability of mutation and contamination is higher
and that the definition of a batch size is not evident.
Batch processes are simple and robust but the only

0167-7799/02/$ see front matter 2002 Elsevier Science Ltd. All rights reserved. PII: S0167-7799(01)01913-8

104

Opinion

TRENDS in Biotechnology Vol.20 No.3 March 2002

Strain construction
Feedback
Fermentation process:
quality control and development
Improvements
and
scale-up

Feedback
Downstream process:
quality control and development

Clinical productions
Quality assurance
and quality control
Manufacturing
Requirements
TRENDS in Biotechnology

Fig. 1. This diagram

shows the interactions
between the different
departments involved in
the development of a
biopharmaceutical
product.

way to reach a high cell density is the fed-batch mode,

which is more complex but allows the metabolism of
the strain to be controlled. For S. cerevisiae, the high
concentration of glucose induces the Crabtree effect:
the yeast enters in a fermentative metabolism even if
there is no limitation of oxygen [11]. The metabolism
can be controlled by using a fed-batch method. Feedlimiting the glucose concentration in the broth allows
the cells to be grown to high density in a respiratory
metabolism for S. cerevisiae [12] and avoids the
production of acetate by E. coli [13]. The comparison
of the productivity of the recombinant product per
day is used to choose the most economical process.
The medium can be either completely defined or
complex when it includes peptones. Complex media
are easy to prepare, cheap and fast growing but are
difficult to analyse and variation between batches
can occur. Defined media gives reproducible growth
and the purity in the supernatant for secreted
protein is better than with complex media. In both
cases, the choice of the carbon source is crucial. High
concentration of glucose will lead to the accumulation
of acetate, which reduces biomass yield and
expression levels in E. coli. Glycerol should be
preferred to glucose for E. coli [13]. Although it
represses the P. pastoris alcohol oxidase promotor,
the glycerol is preferably avoided during induction of
Mut+ strains and can be replaced for MutS strains by
sorbitol, for mixed fed-batch with methanol [14].
The use of raw materials of animal origin must be
avoided when possible. When this is not feasible, the
raw materials must be produced in a country free from
transmissible spongiform encephalopathies (TSEs).
Optimization of the pH medium is crucial
especially for yeast-secreting protein because they
can grow in a wide range of pHs. The choice of pH
depends on the stability of the recombinant protein
expressed. Working with Pichia pastoris at low pH
allows protease degradation to be avoided. This yeast
cannot grow at pHs below pH 2.2.
http://tibtech.trends.com

Temperature has an important role in the

solubility of the expressed protein and also on the
productivity. Lowering the temperature from 30 to
25C during the methanol induction phase results in
a fourfold increase in yield production of galactose
oxydase cloned in Pichia pastoris [15].
For fed-batch processes, exponential feeding
allows the cells to be grown at a constant growth
rate, which is good for expression of recombinant
proteins. The dissolved oxygen is often a limiting
factor if a high growth rate is reached. For
methylotrophic yeast, it is crucial that the cells are
maintained in respiratory metabolism to avoid an
excessive accumulation of methanol in the broth,
which is rapidly toxic. It is useful to supplement air
with oxygen gas [16]. Pressure can be increased to
dissolve more oxygen in the medium but it also
increases the concentration of dissolved carbon
dioxide and thus is perhaps not suitable.
A limitation of dissolved oxygen is not always
detrimental to good expression; which is the case for
an enzyme cloned into E. coli in an L -arabinoseinducible vector. The moment for induction of
recombinant protein expression is optimal when the
oxygen uptake rate is maximized but the dissolved
oxygen concentration is zero [17].
The factors affected by scale are the number of
generations, the mutation probability, medium
sterilization, the quality of temperature and pH
regulations, agitation, aeration and pressure. The
best way to prepare the scaling-up of a process is to
first scale-down to the pilot scale of the conditions of
culture that will be used at the final scale of
production [18]. Then, when the scale is increased,
the broth will become more and more heterogeneous.
In large fermentors, oxygen can be depleted in some
area of the reactor [19].
The process must be characterized and validated
by some controls to demonstrate the robustness of the
procedure and thus its reliability and reproducibility.
All the equipment must be validated for good
manufacturing practices, including the qualification
of the installation, validation of cleaning in place
(CIP) and sterilizing in place (SIP), maintenance and
calibration plan [20]. The fermentation protocol can
be improved until clinical phase III, which must be
performed at the final scale of production.
The optimization of a heterologous protein
production starts with the construction of a
well-designed recombinant strain. The fermentation
development strategy involves the knowledge of the
behaviour of strain and of the recombinant protein.
The fermentation process is developed taking into
account the feedback from the DSP development, the
final scale of production and the good manufacturing
practice requirements. Even though there is a trend
to standardize as much as possible of the
fermentation processes, it has been shown that even
subtle changes would powerfully increase the
productivity. The optimization of the fermentation

Opinion

TRENDS in Biotechnology Vol.20 No.3 March 2002

process should be performed using a two tiered

approach. First the well-known avenues described in
this article and others allow a rational and scientific
optimization. In a second approach the study of the
property of the expressed protein should suggest
more unusual parameters to be assayed.
References
1 Posno, M. et al. (1991) Incompatibility of
Lactobacillus vectors with replicons derived from
small cryptic Lactobacillus plasmids and
segregational instability of the introduced
vectors. Appl. Environ. Microbiol. 57, 18221828
2 Futcher, A.B. and Cox, B.S. (1984) Copy number
and the stability of 2-m circle based artificial
plasmids of Saccharomyces cerevisiae.
J. Biotechnol. 157, 283290
3 Leers, R.J. et al. (1992) Structural and functional
analysis of two cryptic plasmids from
Lactobacillus pentosus MD353 and Lactobacillus
plantarum ATCC 8014. Mol. Gen. Genet.
234, 265274
4 Koizumi, J-I. et al. (1985) Effects of the
temperature and the dilution rate on the copy
number of recombinant plasmid in continuous
culture of Bacillus staerothermophilus (pLP11).
Biotechnol. Bioeng. 27, 721728
5 Reinikainen, P. and Virkajrvi, I. (1989)
Escherichia coli growth and plasmid copy
numbers in continuous cultivations.
Biotechnol. Lett. 11, 225230

Lara Marks and Emmett Power

It takes nearly ten years to get a drug through the discovery and development
pipeline and onto the market; most of this time is spent in the clinical phase.
Clinical development times vary widely from drug to drug, but a drug typically
spends just over 6 years going through clinical trials and regulatory processes.
At least 3 years of this time is spent recruiting patients. Every month by which
the development process can be shortened is worth US $25 million in additional
income for the average drug. Can the recruitment time be shortened?

In the biopharmaceutical industry, time matters more

than money. A drug typically costs US $75 million
over the 6 years that it takes to go through clinical
trials and approval, which amounts to a development
cost of just under US $1 million a month. However,
the income generated by a typical drug marketed by a
http://tibtech.trends.com

For example, adjunction of a cofactor or substrate

can stabilize an enzyme and increase the final
productivity. In this area of research, the imagination
and creativity of the scientist can achieve outstanding
results that are well out of reach of the best
multifactorial experiment plan.

6 Pinches, A. et al. (1985) Growth, plasmid stability

and -amylase production in batch fermentation
using a recombinant Bacillus subtilis strain.
Biotechnol. Lett. 7, 621626
7 Kumar, P.K.R. et al. (1991) Strategies for
improving plasmid stability in genetically
modified bacteria in bioreactors.
Trends Biotechnol. 9, 279284
8 Gerdes, K. (1988) The parB (hok/sok) locus of
plasmid R1: a general purpose plasmid
stabilization system. Biotechnology 6, 14021405
9 Scheirlinck, T. et al. (1989) Integration and
expression of -amylase and endoglucanase genes
in the Lactobacillus plantarum chromosome.
Appl. Environ. Microbiol. 55, 21302137
10 Murphy, T.D. (1977) Design and analysis of
industrials experiments. Chem. Eng. 6, 168182
11 Kpelli, O. (1986) Regulation of carbon
metabolism in Saccharomyces cerevisiae and
related yeasts. Adv. Microb. Physiol. 28, 181208
12 Mendoza-Vega, O. et al. (1994) Industrial
production of heterologous proteins by fed-batch
cultures of the yeast Saccharomyces cerevisiae.
FEMS Microbiol. Rev. 15, 369410

Using technology to
address recruitment
issues in the clinical
trial process

105

13 Lee, S.Y. (1996) High cell-density culture of

Escherichia coli. Trends Biotechnol. 14, 98105
14 Streekrishna, K. et al. (1997) Strategies for
optimal synthesis and secretion of heterologous
proteins in the methylotrophic yeast
Pichia pastoris. Gene 190, 5562
15 Whittaker, M.M. and Whittaker, J.W. (2000)
Expression of recombinant galactose oxidase by
Pichia pastoris. Protein Expr. Purif. 20, 105111
16 DAnjou, M.C. and Daugulis, A.J. (2000)
Mixed-feed exponential feeding for fed-batch
culture of recombinant methylotrophic yeast.
Biotechnol. Lett. 22, 341346
17 Doig, S.D. et al. (2001) Large scale production of
cyclohexanone monooxygenase from Escherichia
coli TOP10 pQR239. Enzyme Microb. Technol.
28, 265274
18 Kwanmin, J.J. (1989) Scale-down techniques for
fermentation. Biopharm. 2, 3039
19 Enfors, S. et al. (2001) Physiological responses to
mixing in large scale bioreactors. J. Biotechnol.
85, 175185
20 Naglak, T.J. et al. (1994) Validation of
fermentation processes. Biopharm. 7, 2836

large biopharmaceutical company is approximately

US $300 million a year (US $25 million a month).
Given this cost-to-benefit ratio, biopharmaceutical
companies tend to focus on moving drugs faster
through the development process rather than
reducing the cost of development, which is easier
said than done. The clinical cycle is a long,
complicated process that has been developed over a
long period. The clinical process involves several
participants including regulators, ethics committees
and clinical investigators, all of whom have to be
convinced that any change to the process makes
sense. This can be challenging and so it is not
surprising that biopharmaceutical companies often
opt to maintain the status quo. Despite this, the
rewards for incremental changes to the process can
be immense. A lack of faith in their ability to execute
these changes flawlessly, and the fear of the
consequences of poor execution, lies behind the
failure of biopharmaceutical companies to seize
these opportunities.
The clinical process itself involves several
complex sub-processes and several strategies can be
adopted to decrease clinical development times.
From our studies, we have concluded that an
optimal risk-adverse strategy for addressing issues
of cycle times is to address each sub-process
individually. The optimal risk-avoidance strategy
begins with those sub-processes that deliver the
greatest benefit for the least effort over the shortest

0167-7799/02/$ see front matter 2002 Elsevier Science Ltd. All rights reserved. PII: S0167-7799(01)01881-9