Clemens Posten

Integrated Bioprocess Engineering

Also of Interest
Biomimetic Nanotechnology
Senses and Movement
Mueller, 2017
ISBN 978-3-11-037914-3, e-ISBN 978-3-11-037916-7

New-Generation Bioinorganic Complexes

Jastrzab, Tylkowski (Eds.), 2016
ISBN 978-3-11-034891-0, e-ISBN 978-3-11-034890-3

Microbial Applications.
Recent Advancements and Future Developments
Kumar Gupta, Zeilinger, Ferreira Filho, Carmen
Durán-Dominguez-de-Bazua, Purchase (Eds.), 2016
ISBN 978-3-11-041220-8, e-ISBN 978-3-11-041278-9

Pharmaceutical Chemistry
Vol. 1: Drug Design and Action
Campos Rosa, Camacho Quesada, 2017
ISBN 978-3-11-052836-7, e-ISBN 978-3-11-052848-0

Pharmaceutical Chemistry
Vol. 2: Drugs and Their Biological Targets
Campos Rosa, Camacho Quesada, 2017
ISBN 978-3-11-052851-0, e-ISBN 978-3-11-052860-2
Clemens Posten

Integrated
Bioprocess
Engineering
|
Author
Prof. Dr.-Ing. Clemens Posten
Karlsruher Institut für Technologie (KIT)
Inst. Bioprocess Engineering
Fritz-Haber-Weg 2
76131 Karlsruhe
clemens.posten@kit.edu

ISBN 978-3-11-031538-7
e-ISBN (PDF) 978-3-11-031539-4
e-ISBN (EPUB) 978-3-11-038200-6

Library of Congress Control Number: 2018934812

Bibliographic information published by the Deutsche Nationalbibliothek

The Deutsche Nationalbibliothek lists this publication in the Deutsche Nationalbibliografie;
detailed bibliographic data are available on the Internet at http://dnb.dnb.de.

© 2018 Walter de Gruyter GmbH, Berlin/Boston

Cover image: Clemens Posten
Typesetting: le-tex publishing services GmbH, Leipzig
Printing and binding: CPI books GmbH, Leck
♾ Printed on acid-free paper
Printed in Germany

www.degruyter.com
Preface – a short pitch for this book
Modern biotechnology has made tremendous and amazing progress. The driving force
is the desire to deliver products for urgent societal needs. This development has been
rightly respected and has found its echo in public perception in general and in teach-
ing especially. However, technical processes are necessary to make these benefits a
reality. Bioprocess engineering has stayed a bit in the shade; reason enough to write a
book not only as tour of the possibilities but with also with a didactic claim. Already
during the first leaf through you will find formulas that link a process and its physi-
cal or chemical basis, data and simulation plots that give a feeling for real production
values, and numerous pictures to underpin a visual understanding. The intention in
writing this book was to do a bit more on the didactical level than can usually be found
in textbooks on bioprocess engineering and certainly than the fragmentary informa-
tion that the internet can deliver. Bioprocess engineering stands in the tension be-
tween biology, technical feasibility and societal demands. New concepts like the ‘cell
factory’ are answers to this challenge. The bioprocess engineer has to integrate these
aspects into the design of modern and successful production processes. He must be
prepared and sensitized for it, making the programmatic title “Integrated Bioprocess
Engineering” sensible.
Dear reader of this preface: Your most urgent question at the moment may be
whether this book is suitable and enlightening for you. In fact, it is designed for un-
dergraduate and graduate students of bioengineering. Previous knowledge in reaction
engineering and/or microbiology is helpful, but not absolutely necessary. This book
is therefore also aimed at newcomers from biology and chemical engineering. Missing
background knowledge can easily be gained from other information sources. Dear stu-
dent, the book is organized along a thematic thread and goes sometimes into depth,
which does not intuitively follow the Bachelor level. Don’t be unsettled, but feel en-
couraged to study these aspects again during your Masters courses. This book wants to
be a faithful companion through the whole study. Why not read this book as a teacher,
journalist, politician or interested layman? You will get a dense background in the fast
emerging field of bioprocess engineering to support your own opinion and qualified
decisions. The natural scientist or engineer already in the profession will also find
many new facts and approaches. Although the content of the book is not intended as
reference book for ‘experts’, it gives the framework to mentally integrate your own ex-
periences, structure your own knowledge, and become a more effective and creative
professional.
What is the content of the book, how is it organized, and how can you person-
ally access the book? The chapters follow the workflow of process development from
selection of the production organism and media design, to reactor operation and pro-
cess policy. Some chapters cover crosscutting issues like measurement and modeling.
The chapters are built on each other, so working through the book in this consecutive

https://doi.org/10.1515/9783110315394-201
VI | Preface – a short pitch for this book

order is strongly recommended. Cornerstones of the material are process examples

from flow sheets down to evaluated data. The selection of the example processes re-
flects mainly the topic in the respective chapter. Yeast production is a classic and well
understood process. This circumstance allows an accurate and instructive representa-
tion from microbiology to process management and product design, so it is chosen as
running example through the chapters. On the other hand, phototrophic microalgae
production is a new and dynamically developing field, which has a separate chapter
dedicated to it. There are many different processes in practical use, so that it is not
possible to show even the most important of them. The book is not as inclusive as
possible, but tries to convey the concepts of fast and effective new processes in or-
der to understand, get a feeling for, and improve them. Besides the thread of process
development, the mediation of these concepts is the real didactic structure, forming
braces between the chapters. To make it visible, each chapter concludes with a para-
graph summarizing and highlighting such concepts of structured thinking including
different aspects of integration.
If there are questions unanswered, if you have ideas for other process examples,
or even if you find ambiguous formulations, pleased don’t hesitate to contact me. Now
enjoy reading, accompanied by a quote of Louis Pasteur, one of the greatest forefathers
of biotechnology: “Fortune favors the prepared mind.” My hope is that this book will
help the esteemed reader to follow his spirit a little bit.

Clemens Posten, Karlsruhe, Germany

New Year’s Day 2018
Contents
Preface – a short pitch for this book | V

1 Introduction – a thread through this book | 1

1.1 Motivation – window shopping for biotechnological products | 1
1.2 Bioprocess engineering – attempt at a definition | 7
1.3 Yeast production – a classic but instructive process | 8
1.4 The three columns of bioprocess engineering – what bioprocesses have
in common | 11
1.5 Integration of sciences – acquiring knowledge on demand | 13
1.6 Characteristics of living cells – unique selling points from an
engineering view | 14
1.7 Exercises, questions and suggestions | 16

2 Biosystems – microorganisms and other biocatalysts | 17

2.1 Motivation – a first bioengineering view of microbiology | 18
2.2 Diversity of biosystems – appearance in technical environments | 19
2.2.1 Enzymes – the universal toolbox for biocatalysis | 19
2.2.2 Bacteria – organisms for all seasons | 21
2.2.3 Archaea – the underestimated extremists | 23
2.2.4 Yeast – the workhorse of bioprocesses | 24
2.2.5 Filamentous fungi – the broad-spectrum chemists | 26
2.2.6 Microalgae – the solar cell factory | 27
2.2.7 Plant cells – slow but resourceful | 29
2.2.8 Animal cells – masters of protein decoration | 31
2.2.9 Organized communities – making rich booty as a consortium | 32
2.3 Cultivation conditions – environmental parameters to care about | 33
2.3.1 Primary nutritional groups | 33
2.3.2 Temperature | 34
2.3.3 Acidity and basicity | 37
2.3.4 High concentrations | 38
2.4 Bringing microorganisms to work – the concept of a ‘cell factory’ | 39

3 Media – supplying microorganisms with a comfortable environment and

building blocks for growth | 41
3.1 Media – the material basis of the process | 41
3.2 Media design – starting from scratch | 42
3.3 Practical Application – a discussion of recipes from references | 48
3.4 Complex and technical media | 53
3.5 Additional aspects – mutual interactions between cells and
medium | 58
VIII | Contents

3.6 The Good, the bad and the ugly – microorganisms as products | 61
3.7 Exercises, questions and suggestions | 64

4 Kinetics – finding quantities for bioprocess reactions | 66

4.1 Kinetics – the scaffold of reaction engineering | 66
4.2 Enzymes as basic components – determining kinetics | 68
4.3 The specific growth rate – describing growth by numbers | 76
4.4 The yield coefficient – combining substrate turnover and growth | 82
4.5 The batch process – the simplest form of a bioprocess | 86
4.6 Exercises, questions and suggestions | 88

5 Bioreactors – designing a home for the bioreaction | 92

5.1 Not only a vessel – for what a bioreactor is needed and what it
can be | 92
5.2 Revisiting the stirred tank reactor – the most important issues to learn
from chemical engineering | 94
5.3 Pneumatically driven bioreactors – a soft way of energy transfer | 104
5.4 Other types of bioreactors – translating demands into design | 108
5.5 Gas transfer – supplying microorganisms with gaseous
compounds | 116
5.5.1 Transport processes – the journey of gases through the reactor | 117
5.5.2 Measuring gas transfer | 123
5.5.3 Alternative means of gas supply | 125
5.6 Exercises, questions and suggestions | 127

6 Not always so simple – the batch process reconsidered | 129

6.1 Formal kinetics – extrapolation from enzymatic reactions to cell
growth | 129
6.2 Looking a step deeper – estimating aerobic growth yields from
metabolic fluxes | 132
6.3 Aerobic growth – case study of heterotrophic plant cells | 134
6.4 Looking a step deeper – estimating anaerobic product yields from
metabolic fluxes | 137
6.5 Anaerobic batch culture – case study of ethanol production | 138
6.6 Running example of yeast production – growth in batch
processes | 142
6.7 Primary metabolites – products directly taken from central
metabolism | 145
6.8 Serving mankind with basic needs – products needed in large
amounts | 148
6.9 Conclusions for batch processes – dealing with pros and cons | 153
6.10 Questions and suggestions | 155
 Contents | IX

7 Little by little one goes far – the fed-batch process | 157

7.1 Setting up process equations – the formal deduction of a suitable
feeding strategy | 158
7.2 Running example yeast production – an industrial feeding
strategy | 161
7.2.1 Deviations from exponential – controlling physiology and obeying
technical constraints | 161
7.2.2 The first step of downstream processing – relationship between cell
separation and cell physiology | 165
7.3 Production of secondary metabolites – case study of penicillin
production | 168
7.4 Production of recombinant proteins – the cell factory | 173
7.4.1 Different ways to get hold of desired proteins – an overview of protein
expression systems | 174
7.4.2 High density cultivation – a specific fed-batch process for recombinant
protein production | 176
7.4.3 Production of pharmaceutical proteins with Kamagataella pfaffii –
employment of a high performance factory | 178
7.5 Conclusions for fed-batch processes – dealing with strengths and
weaknesses | 180
7.6 Questions and suggestions | 181

8 Microalgae – the solar cell factory | 182

8.1 Becoming curious – current research promises | 182
8.2 Choosing from diversity – widely used microalgae for commercial
bioprocesses | 183
8.3 Physiological principles – the unique features of the microalgal
cell | 186
8.4 Light – the little bit different energy source | 191
8.5 Media for microalgae – a clear matter of fact | 194
8.6 Kinetics of photobioprocesses – how microalgae make use of the
nonmiscible substrate | 197
8.7 Photobioreactors – the interface between algae and sunlight | 203
8.7.1 Starting from natural conditions – lakes and open ponds | 203
8.7.2 Flat plate reactors – the bubble column in an ‘enlightened’
guise | 206
8.7.3 Tubular reactors | 208
8.7.4 New developments and ideas | 210
8.8 Products and processes | 211
8.8.1 Products for food supplements, feed and aquaculture | 211
8.8.2 High value products | 213
X | Contents

8.8.3 Conversion of photon energy to chemical energy – producing fuels

by microalgae | 216
8.9 Outlook | 218
8.10 Questions and suggestions | 219

9 Continuously operating bioprocesses – production under steady state

conditions | 220
9.1 Setting up stationary balance equations – a good start in understanding
process behavior | 221
9.2 Ethanol production in a continuous process –
the window of operation | 226
9.3 Enzymatic processes – a simple but effective example for continuous
bioprocess operation | 229
9.4 Biogas production via anaerobic digestion | 231
9.5 From minerals and particles – technical concepts for using
chemolithotrophic microorganisms | 236
9.6 Complex molecular structures – how they influence the process | 243
9.7 Current discussion about continuous processes – motivation and
obstacles | 247
9.8 Questions and suggestions | 248

10 Measuring principles – how to put an end to the blind flight | 250

10.1 Only a look through the keyhole – what measurement means for process
understanding and optimization | 250
10.2 State of the art – overview of measurement at a standard reactor | 250
10.3 Physical parameters – adaptation from other fields of process
technology | 253
10.4 Chemical parameters – employment of electrochemical effects | 254
10.5 Measuring biomass – the great unknown ‘X’ | 262
10.6 Compiling a construction kit for sensors – a general approach for
biological sensors | 268
10.7 Questions and suggestions | 271

11 The practice of fermentation – a step by step guide through the

workflow | 272
11.1 Structuring the reactor and the cultivation – zooming in from principle
understanding to visibility of process details | 272
11.2 The reactor – a complex assembly of parts and modules | 273
11.3 Structuring the cultivation process – from planning to final data
acquisition | 275
11.4 Media preparation | 277
11.5 Preculture preparation | 278
 Contents | XI

11.6 Reactor installation and setup | 279

11.6.1 Fermentation vessel setup | 280
11.6.2 Peripheral devices | 281
11.6.3 Sensor installation | 282
11.6.4 Autoclaving the bioreactor | 283
11.7 Managing the actual cultivation process | 287
11.8 Integration on the process level – how to include the fermentation into a
whole process chain | 292
11.9 Questions and suggestions | 293

12 Modeling – art and handcraft of mathematically describing

bioprocesses | 295
12.1 Modeling – what it might be and what it is good for | 295
12.2 Predator-prey model | 299
12.3 Linear networks – finding stationary flux equations | 302
12.4 Aerobic growth – setting up a first general model | 304
12.5 Anaerobic growth – traps and pitfalls | 307
12.6 Back to baker’s yeast – more degrees of freedom | 310
12.7 Back to microalgae – a more spatial and hierarchical structure for
subsystem definition | 315
12.8 Integration into society – the final proof of meaning | 319
12.9 Suggestions | 322

Further Reading – still curious? | 323

Acknowledgments – dedicated to all the people who supported the compilation of

this book | 327

Copyrights – pictures provided with courtesy and accepted with thanks | 333

Index | 335
1 Introduction – a thread through this book
Biotechnology is a fascinating science and an invaluable weapon in the battle against
hunger and diseases. It uses the capabilities of enzymes, microorganisms and isolated
cells that produce complex molecules to bring comfort to our daily lives. Success sto-
ries about groundbreaking techniques in genetic engineering and the discovery of new
bioactive compounds are at the center of public interest. Integrated bioprocess engi-
neering provides the means to go from ideas to actually achievable products for the
benefit of our society. This is not an easy path. The challenge can only be approached
by sound application of engineering sciences and smart use of biological capabilities.
The chapters of this book follow the workflow during usual bioprocess design and
operation, from lab work through fermentation to the final product. On the way, stan-
dard unit operations such as bioreactor design, including mixing and aeration, are
addressed. This is achieved by fundamental chapters giving descriptions of state-of-
the-art techniques and leading the reader through the jungle of technical calculations.
Beyond this ‘hardcore’ process engineering, concluding paragraphs raise additional
questions and suggestions meant to strengthen the ability to take the broader view,
which is essential for further success.
In several case studies special processes will be introduced. Different aspects
of the interplay between cell physiology and the process conditions become clear
for these examples. Here, know-how is complemented by know-why for these special
cases. This is meant to allow for knowledge transfer from basic approaches to the huge
diversity of other already existing processes, and to allow the reader to find ideas for
their own projects. Further examples are given in smaller formats as an overview of
different solutions for a given problem.
Integration takes place on different levels from direct coupling of process steps
to societal aspects. The examples for integrative aspects given in the previously men-
tioned chapters are summarized and generalized in specifically dedicated paragraphs.
Frameworks and concepts of the current way of thinking in science and industry are
highlighted and outlined. Finally, exercises provide an opportunity to reinforce and
train the teaching contents, as well as to encourage further investigations.

1.1 Motivation – window shopping for biotechnological products

People encounter more and more biotechnological products in everyday life, whether
they know it or not. Dairy products like yoghurt or kefir are traditional examples. Fer-
mented foods are found in many cultures all over the world. Wine and beer have con-
sumed for millennia thanks to the fermentative activity of yeast. Furthermore, yeast
lets dough rise. Acetic acid has been known since antiquity for the acidification and
conservation of foods. Since the Middle Ages it has been produced by the so-called

https://doi.org/10.1515/9783110315394-001
2 | 1 Introduction – a thread through this book

Fig. 1.1: Traditional food: (a) vinegar (aceto balsamico), one of the oldest biotechnological foods,
presented here in modern lifestyle appearance (© Meine Pestoria); (b) cheese is a traditional food
as well. Scrubbing, basically inoculation with selected microorganisms, supports the formation of
the rind during cheese ripening (© Schönegger Käse-Alm).

Orleans method, an example for surface fermentation. This provides the acid bacte-
ria for oxidation of ethanol (wine) to acetate. Later it was intensified (by increasing
the air/water interface) by letting the suspension trickle over beech wood chips. For
technical usage, acetic acid production even by modern fermentation processes is not
competitive with chemical production. That does not hold for vinegar, a good example
of integration into society, where public perception is an important issue, especially
in the food area Figure 1.1.
Convenience food is commercially processed food that includes biotechnological
steps or ingredients to optimize taste, smell or ease of consumption. The flavor en-
hancer glutamate is the most prominent example and is present in many different
packaged foods. It has been produced in a direct fermentation process since the 1950s
by the bacterium Corynebacterium glutamicum. Today, total world production of this
amino acid is estimated to be two million tons per year. Less well known is the polysac-
charide xanthan. It is present as food thickener and gelling agent in many soups,
dressings and ice creams. It is produced by the bacterium Xanthomonas campestris.
Citric acid is not only used in detergents to dissolve limescale but also for acidifica-
tion of fruit juices. Its production with Aspergillus niger is one of the most important
biotechnological processes with respect to market volume Figure 1.2.
 1.1 Motivation – window shopping for biotechnological products | 3

Fig. 1.2: Convenience food: (a) many instant soups, here the famous Asian ramen soup, contain
glutamate or yeast extract as flavor enhancers; (b) sauces like salsa and dips get their physical
consistency from xanthan as food thickener and contain acetate, citrate, or glucose fructose syrup for
better taste.

A modern aspect in our daily diet is the regular consumption of food supplements and
functional foods. An example of living cells as a product is found in probiotics. They
contain live bacteria and yeasts that are meant to be good for human health, espe-
cially the digestive system. However, the medical effect is not fully assessed. Polyun-
saturated fatty acids (PUFA) and antioxidants are typical food supplements that are
increasingly derived from microalgae, thus relieving the limited resource fish oil. Vi-
tamin C is present in multivitamin products or in high doses in effervescent tablets.
Each supermarket has a wide choice of different formulations of vitamins Figure 1.3.
Cosmetics is another field for biotechnological products. Many face creams con-
tain hyaluronic acid (HA). This polysaccharide is widely distributed in the extracellu-
lar matrix of animal connective and epithelial tissues, where it fulfills many chemical
and physical functions. Its application as a cosmetic agent promises skin tightening.
HA is still extracted from cockscombs. Growing ethical thinking on the producer and
the consumer side has led to a biotechnological process, where HA is produced by fil-
amentous fungi. This ‘bio-hyaluronic acid’ captures more and more market share. In
this case, production by plants is not an alternative as for other compounds like lipids
or amino acids. A protective agent for microorganisms is ectoine, a low molecular nat-
ural compound. It is present in many microorganism that can stand high temperature
4 | 1 Introduction – a thread through this book

Supplement Facts Serving Size: 1.8 oz

Vitamin A 100 μg
Vitamin C 16 mg
Vitamin E 2 mg
Vitamin D3 1.6 μg
Vitamin B12 0.7 μg
Vitamin K 5 μg
Vitamin B12 0.5 μg
Thiamin (B1) 0.5 mg
Riboflavin (B2) 6 mg
Vitamin B6 0.7 mg
Pantothenate 1.3 mg
Folic acid 106 μg
Bion 12 μg
Niacin 5 mg
(a) (b)

Fig. 1.3: Food supplements and functional food: (a) living bacteria are the biotechnological basis of
fermented dairy products and are nowadays specifically selected and designed for use in functional
foods, here a probiotic milk drink; (b) vitamins as supplements to the usual diet have become stan-
dard in many countries. Which of the vitamins listed on the label of a multivitamin preparation are
biotechnologically produced.

and salinity. It is used as an active ingredient in skincare products, protecting the skin
against ultraviolet (UV) light and mucous membranes against dryness Figure 1.4.
Biotechnologically produced drugs for pharmacological application are of spe-
cific importance, as in most cases they cannot be manufactured by any other means
or substituted readily by other substances. Penicillin is the classic example, which
has extremely beneficially contributed to human health. Nowadays different antibi-
otics are available in any pharmacy. Bad compliance has led to spreading of multidrug
resistant pathogens, an example where integration into society creates a drawback.
A more modern product in healthcare is the peptide hormone insulin, which is also
known to everybody. Other highly effective vaccines or antibodies are available only
in hospitals Figure 1.5.
The portfolio of bioproducts in the areas of food and pharmaceuticals have long
been exceeded by products for practical use. Many products can be found in ev-
ery household and are available in every supermarket. Technical enzymes like pro-
teases or lipases are important active compounds in washing powder and cleaning
agents. Another group of products is detergents, an application of surface active
molecules (surfactants). Biotechnologically produced and biodegradable surfactants
(e.g., rhamnolipids) are labeled as ‘biotensides’ and are produced by microbial fer-
 1.1 Motivation – window shopping for biotechnological products | 5

Fig. 1.4: Cosmetics: (a) product for skincare based on hyaluronic acid written as ‘biohyaluronic acid’
when produced by microorganisms; (b) nasal spray with ectoine to prevent drying out of the nasal
mucous membrane in winter when heated air in the living room is very dry.

Fig. 1.5: Pharmaceutical products: (a) penicillin is still the most powerful weapon against bacterial
infections; (b) insulin was the first medical drug made by genetically engineered bacteria. Since
then it has helped millions of people live with diabetes.
6 | 1 Introduction – a thread through this book

Fig. 1.6: Products for use in the household: (a) washing powder contains different enzymes like pro-
teases; (b) a relatively new product is biotensides in environmentally friendly household cleaners.

Fig. 1.7: Technical products: (a) polylactide is one of the current biobased plastics and a major mate-
rial for 3D printing, here employed to form a globe; (b) vehicles can be fueled in many countries with
pure bioethanol or with ethanol as a fuel additive.
 1.2 Bioprocess engineering – attempt at a definition | 7

mentation or enzymatic catalysis of plant derived oils. Especially for outdoor appli-
cations, they are more environmentally friendly than surfactants based on fossil oils
Figure 1.6.
Beyond food, cosmetics and medicine, bioproducts have acquired their place in
technical applications. Sodium gluconate is used as aggregate in concrete mixes and
acts as set retarder to prevent cracking during fast curing. Metal surface treatment
is another application. Great hopes are placed on biobased, biodegradable and bio-
compatible plastics. A success story is polylactides (polylactic acid, PLA), for which
the lactic acid is produced by fermentation. PLA blends are already in use in different
specific applications such as packaging (especially deep drawn products like yoghurt
pots and coffee pads), small parts of technical appliances, or as fibers for technical
fabrics. However, the most significant biotechnological product in terms of volume is
ethanol. Even at the petrol station we can make a find. With bioethanol, biotechnol-
ogy made the step into the energy market. The huge need for sugar cane in competition
with its use as food however makes this process controversial Figure 1.7.

1.2 Bioprocess engineering – attempt at a definition

The bioprocess engineer deals with living material employed in technical processes.
This basic view states two characteristics as the heart of the activity field. However, it is
not intuitively clear exactly what ‘bio’ means and exactly what kind of ‘process’ is the
target. The process or chemical engineer deals with the processing of raw materials
that are being converted into added value products, employing physical, chemical,
and biological means. The characteristic view is to see the materials as a more or less
unstructured flow of substances. This delimits the process engineer from mechanical
engineering handling of single workpieces. It also excludes medical engineering or
agricultural technologies. Manipulating single plants or cells is therefore not a typical
task of a process engineer. However, during a process molecules, particles, or cells are
definitely and intentionally altered.
The term ‘bio’ assigns a role to the bioengineer in the larger field of biotechnology.
The Organisation for Economic Co-operation and Development (OECD) gives a defini-
tion of biotechnology (biotechnology – OECD Factbook 2011–2012-OECD Library):

“The application of science and technology to living organisms, as well as parts, products and models
thereof, to alter living or non-living materials for the production of knowledge, goods and services.”

This definition can be analyzed further. In an older version only microorganisms were
mentioned, whereas now plants and animals are increasingly in the focus. Under
‘parts of cells’ things such as enzymes could be understood. It is also worth stating
that the ‘production of knowledge’ as such is a kind of technology, as long as it is a
8 | 1 Introduction – a thread through this book

targeted process using biological and engineering principles. So, bioprocess engineer-
ing is not only ‘running the process’ but also has to apply the tools and paradigms of
engineering to understand and design processes, and not only production processes.
This definition of the OECD covers all modern biotechnology but also many tra-
ditional or borderline activities. To specify this further, the definition is accompanied
by a list based definition, where item IV states:
“Process biotechnology techniques: Fermentation using bioreactors, bioprocess-
ing, bioleaching, biopulping, biobleaching, biodesulphurisation, bioremediation,
biofiltration and phytoremediation.”
This is a collection of technical means (here only the bioreactor) and different
fields of application. In this book we will discuss specific processes and general meth-
ods going beyond these strict technical aspects. The work of a bioengineer is not lim-
ited to large scale operations as suggested by the definition. Zooming into a cell shows
‘processes’ as well, which can be addressed and investigated by an engineer’s way of
thinking and their specific methods. This includes transport and reaction phenomena
to be described by kinetics, balances and thermodynamics familiar to chemical engi-
neering at the large scale. In fact, chemical engineers were amongst the first drivers
of metabolic flux analysis. Nevertheless, designing of bioprocesses is the core activity
of the bioprocess engineer.

1.3 Yeast production – a classic but instructive process

To get a first insight into the structure of bioprocesses, a closer look to yeast production
will help. The demand for baker’s yeast is directly associated with the need for bread,
making the production a rapidly increasing industry. Growth rates are estimated to
be more than 1% in developed countries and 10% in developing countries. Neverthe-
less, yeast is a cheap product compared to other biotechnological products, making
an optimized process necessary with respect to cost efficiency.
A flow chart is given in Figure 1.8. The workflow starts in the lab where single yeast
cultures are stored in frozen vials or agar slants. Yeast companies hold several strains
in their strain collection to be employed for different regions and different types of
bread. For sugar bread, where the dough may rise for several days, the yeast cells have
to be robust against high osmotic pressure, in contrast e.g., to baguette, where short
term high fermentative activity is needed. The strains originate in best cases from one
single cell. Specific institutes, the yeast banks, host such cell lines for many breweries
and baker’s yeast production companies. The selected strain is further propagated by
transfer into a shaking flask. This can be done by a platinum loop or ‘eye’, from which
this operation has its name ‘inoculum’ (from the Latin oculus, meaning eye). Further
propagation goes through vessels with increasing size up to a final 30 L cultivation ves-
sel. The purpose of this part of the lab work, the seed production, is to provide biomass
as inoculum for production, commercial fermenters, or bioreactors. This sequence of
 1.3 Yeast production – a classic but instructive process | 9

inoculum

cooler
seed reactor producon
reactor
medium
reservoir

disk stack
separator vessel vacuum drum compressed
yeast cream filter yeast

Fig. 1.8: Flow chart of a yeast production process. The pure strain in the inoculum is mixed with the
medium to start the seed reactor. Stepwise scale up leads to the final production reactor. After cool-
ing the yeast is concentrated in a special centrifuge, the disk stack separator, giving yeast cream as
the first product. Production of compressed yeast for the supermarket requires an additional filtra-
tion step.

cultivations is in industry often called the ‘seed train’. ‘Seed yeast’ is produced under
sterile and/or aseptic conditions. Seed yeast cream is separated and stored in seed
yeast cream tanks, also called stock fermenters, and inoculated in portions into sev-
eral production fermenters.
Other work items in the lab include analysis and preparation of the medium, to
supply the cells with the necessary amounts of sugar and other nutrients. Molasses,
a sugar containing waste stream from sugar production, is used in yeast production
for this purpose; see Chapter 3 concerning media design. Molasses preparation is a
major step in the upstream process of yeast production. In order to have a pumpable
medium, molasses have to be diluted with hot water. Diluted molasses will be ster-
ilized continuously under pressure, stored in clean or sterilized molasses tanks, and
fed into fermenters during yeast propagation. In this case, water is treated with chlo-
rine. In some factories diluted molasses will be acidified, sedimentation occurs, and
molasses will be clarified continuously by decanters. The choice of these operation
steps depends on the quality and source of the molasses (beat or cane). With the final
contamination check, the work in the lab is completed. This part of the production
process, delivering the seed, is generally for bioprocesses called ‘upstream’ process-
ing.
The next steps take place in bioreactors located in the production area. Further
yeast propagation occurs in bioreactors of progressively larger volumes VR by a factor
of about 10. The index R means the working volume of the reactor. These large biore-
actors are called in technical jargon ‘fermenters’. Temperature, pH value and medium
10 | 1 Introduction – a thread through this book

concentration are strictly controlled. The pH value is kept at 4.5–5.5, which is quite
acid compared to the optimum for bacteria. Together with the potential of yeast to
produce ethanol, this is considered as one of the reasons why humans were able to
produce yeast despite the contamination risks. The temperature is kept between 30 °C
and 35 °C.
The first steps, called seed fermentation, are kept under aseptic conditions, ex-
cluding other microorganisms from the reactor. The first product is the pure yeast that
can be either sold or used to inoculate the next stage. The amount of yeast transferred
from a small bioreactor into a larger one is still called inoculum, although it has noth-
ing to do with a loop or ‘eye’ any more. Prior to inoculation there can be a concen-
tration step by centrifugation. These seed fermentations ends in 30 m3 scale. In the
final step, named trade or commercial fermentation, which occurs in large scale up
to VR = 300 m3 , the medium is not sterilized due to cost reasons, so the next pro-
duction steps are not necessarily free from contamination, e.g., by ‘wild yeasts’. Such
contaminations are not expected to propagate for the duration of one or two trade
fermentations. Nevertheless, cleaning of equipment, steaming of pipes and tanks as
well as air filtering is practiced to ensure aseptic conditions. These central steps during
production, where basically biomass is propagated, form the ‘bioreaction’ stage.
At this point we have to think about measuring the amount of yeast produced. This
can be done by taking a sample and counting the cells, or filtrating the sample and
weighing the ‘wet cells’. Both methods are problematic as cells are not all the same
size and contain different amounts of water, between 80–90%. In addition, there is
gusset water between the cells. For engineering and selling purposes a reliable mea-
sure in terms of mass is needed. This is approached by sampling and drying the sam-
ples according to a standardized protocol. This includes filtration or centrifugation of
a sample and drying it at 100 °C for up to 24 h. After drying, the samples consist only
of the solid parts of the cells and do not contain free or physically bound water. This
measure is referred to as cell dry mass, m X [g]. The amount of yeast in the bioreactor
in terms of concentration is then given by cell dry mass per liquid volume c X [g/L]. In
shaking flasks typical values are around 5 g/L, while in the production reactors 50–
60 g ⋅ L−1 are reached before harvesting.
Yeast out of the trade reactors has a cell dry mass concentration c X of about 6%,
which corresponds to 60 g/L. During the next processing step, conditioning for the
market, the yeast cream is further concentrated by several washing, centrifugation
and filtration steps. A disk stack centrifuge delivers a suspension of about 180 g/L,
which is still pumpable. This ‘yeast cream’ can be sold to industrial bakeries. The vac-
uum drum filtration leads to a filter cake with 30% solids content. This final product,
the compressed yeast, is sold in big blocks for bakeries or in small (traditionally 42 g)
yeast cubes in supermarkets. Product formulation includes application of additives
like oils for better handling. These operations of a bioprocess are in general referred
to as ‘downstream processing’. Vinasse, the liquid phase from the concentration steps,
is still rich in some compounds and can be sold as a byproduct after dehydration.
 1.4 The three columns of bioprocess engineering – what bioprocesses have in common | 11

Each cultivation step occurs within 24 h, inclusive of reactor cleaning, and the
whole chain from the seed to the product occurs in one week. This is to keep a standard
schedule every day and week; a tribute to human working conditions and an example
for operational integration of bioprocesses into the social environment.

1.4 The three columns of bioprocess engineering –

what bioprocesses have in common

A generic bioprocess is shown in Figure 1.9. It contains the most important operations
already known from the yeast production process. These are basically present in most
biotechnological production processes. The chapters of this book follow this generic
process from upstream through bioreaction to the downstream stage. Consequently,
we start by looking at the upstream stage, including to begin with microorganisms,
the unique living entities that make a process a bioprocess. Which of them are prac-
tically employed in bioprocesses, and what are their specific needs and capabilities?
In a broader sense, we also have to deal with the biosystems of cells of higher organ-
isms and enzymes. In the next step, we look at media, like molasses in the case of
yeast. Media designs can be straightforward using balance equations and other ratio-
nal considerations. Nevertheless, a lot of empirical knowledge is still behind the usual
recipes. The thread then runs to the bioreaction stage.
The core element is of course the bioreactor. The functions of the bioreactor in the
bioreaction stage are analyzed and we show how the demands are represented in the
technical design. The functions include supplying the organisms with appropriate en-
vironmental conditions with respect to temperature, nutrient and oxygen concentra-

control
(computer) off-gas
analysis
measurement culvaon
(sensor) (bioreactor)

strain selecon analysis

(microorganism) (samples)

medium design feeding

(feeding vessel) (pump) mixing harvesng
(agitator & motor) aeraon (separator)
(valve)

Upstream Bioreacon Downstream

Fig. 1.9: Flowsheet of a generic bioprocess showing all major parts of a production line. The single
elements are generic insofar as the related function can be taken on by other hardware components
from case to case.
12 | 1 Introduction – a thread through this book

tions, as well as protection against undesired microorganisms. But also humans work-
ing with the reactor have their claims concerning manageability and safety. Technical
means are e.g., agitation, aeration, and heat transfer. However, value creation hap-
pens by growth and product formation. Kinetics and process dynamics are means to
describe and control these processes. Especially, interaction of the microorganisms
with media have to be reconsidered for optimal design of the reaction process. To
control dynamics of a bioprocess regardless of the highly complicated interrelations,
specific process control policies have been developed for bioprocesses. These will be
deduced step by step. Nobody wants to run a cultivation in blind flight. So the next im-
portant point is measurement, a topic that has made much progress in recent years.
Based on respective data, mathematical process modeling is state of the art in sci-
ence and industry. The basic approach to modeling will be given in a separate chapter
(Chapter 12).
Finally, the read thread reaches the downstream stage. In many companies this
section is even located in an area spatially separated from the cultivation facilities.
The centrifuge is a symbol for harvesting, the first step of the downstream processes.
Further steps follow but are not the subject of this book. In case studies the interplay
between the three stages of bioprocesses will become clear. In fact, the different oper-
ations can be directly integrated on the process level, overcoming the concept of unit
operations. On the level of process development, integration means the constructive
cooperation of biologists, chemists, process engineers and basically all other sciences
as depicted in Figure 1.10
The highest duty of the process engineer is to become integrated into the devel-
opment team and to integrate knowledge and skills from people in the upstream and
downstream area into their work and decisions. Biochemical engineering science as

Informaon Flow

Upstream Bioreacon Downstream

screening reactor design separaon
work media flow
physiology cell
genec development disrupon
engineering process formulaon
control Policy

Microbiology, Genecs Process Engineering

Chemistry, Analycs Modeling, Measurement

& Control

Fig. 1.10: The three columns of a bioprocess, namely upstream, bioreaction, and downstream.
The basic steps present in nearly all bioprocesses are also indicated.
 1.5 Integration of sciences – acquiring knowledge on demand | 13

well as practical implementation need an interdisciplinary mindset. While the work-

flow goes from upstream to downstream, information flow in all directions is a neces-
sity. For a tongue-in-cheek example: What a mess it would be if a highly productive
strain was isolated and developed in the lab, and then years later the strain was tested
in commercial media in large scale where strong foaming is defeated with surface ac-
tive antifoam agents, finally leaving the membrane separation unit with a mission im-
possible.

1.5 Integration of sciences – acquiring knowledge on demand

First and foremost, bioprocess engineering is the application of process engineering

means to biological systems. The easiest option is to cherry pick from the pool of unit
operations and devices in process engineering. Here we meet aspects of mechanical
and chemical engineering as well as biotechnology and informatics. Calculations are
in most cases given as correlations based on natural laws and experiences. We have to
carefully check the constraints of the validity of the equations given in references or
the suitability of standard equipment. By definition, process engineering deals with
‘unformed’ materials like powders or suspensions and not with direct manipulation
of single objects. Nevertheless, the focus of bioengineering is to look at biological sys-
tems. Microorganisms of course have their own special qualities, so that adjustment
is necessary to meet the specific demands and to harness their full potential. Biolog-
ical knowledge is available in descriptions of the physiology of organisms including
cell morphology or metabolic pathways. The task of the bioengineer here is to find
reliable interfaces between biology and process conditions. These are in best cases
kinetics and stoichiometry, but could include complex cell adaptation as response to
changing environmental parameters.
To do so, a standardized approach may be followed based on thinking in hierar-
chical levels and problem decomposition. On the process level, upstream, bioreaction,
and downstream have to be investigated as shown above. On the reactor and cultiva-
tion level we can virtually separate gas, liquid, and the biophase and understand their
interactions. Here we meet all fields of mechanical and chemical engineering. Design
criteria for reactors seem to be well developed and quite clear. Nevertheless, the bio-
engineer has to make specific adaptations in geometry and material to meet the cells’
needs e.g., for oxygen supply coupled with low shear stress. On the cell level the task
is to understand cells’ needs and to go further into detail with respect to metabolism.
Step by step, we come closer to basic physical or chemical laws. At the end of all our
attempts additional knowledge is necessary and we must design targeted experiments
to acquire data e.g., for kinetics, material or reaction constants. All the small bricks
collected in this way are interconnected. Integration of sciences means to close the
brackets between the single details and to go upwards again, to finally reach a strong
pyramid.
14 | 1 Introduction – a thread through this book

In fact, all applied sciences are based on natural sciences, namely physics, chem-
istry and parts of biology, where mathematics is the common language for description.
This can be regarded as a pyramid of different layers. On each level specific paradigms,
which are fundamental ways of thinking, have already been developed. This makes
things easier and prevents us going over details that may be not be necessary to con-
sider each time. Terms like energy transfer, diffusion or viscosity have been employed
successfully for decades and are still useful for most of the applications. One task in
understanding and designing processes is the formulation of given natural laws or
commonly employed correlations for given conditions. These include geometrical sit-
uations, given materials, or last but not least, specific qualities of the cells. With fur-
ther progress deeper understanding is necessarily gained by looking at intermolecu-
lar forces or complex intracellular metabolic networks. Breakthroughs in engineering
sciences have often been gained by scrutinizing commonly accepted paradigms. Nev-
ertheless, the art of engineering includes the art of simplification. Things should be
as complex as necessary and as simple as possible for a given purpose. Monod formu-
lated his famous kinetics using only two a priori unknown parameters. This turned
out to describe growth curves fairly precise despite the thousands of enzymatic steps
in the bacterial metabolism.
Bioprocess engineering can also give something back to the understanding of bi-
ological systems. Controlled cultivation in bioreactors keeps the cells in an adapted
and defined state and allows us to precisely measure metabolic fluxes and to identify
intracellular bottlenecks.

1.6 Characteristics of living cells – unique selling points from an

engineering view

Bioprocess engineering delivers platform technologies for a spectrum of microorgan-

isms and classes of products to reduce development time for new processes. Neverthe-
less, for specific products adaptations have to be made. In any case, the employment
of a bioprocess has to be justified against the background of costs and societal needs.
The discussion should start with a look at microorganisms and at what makes them
different from inorganic catalysts. Basic points are shown in Figure 1.11
The most fantastic feature of living cells is self-reproduction, being the basis for
life in general. No classic catalyst has this ability. Employing microorganisms for pro-
duction purposes means to employ self-reproduction. A great diversity of microorgan-
isms is available as biocatalysts for different purposes. As engineers we have to pro-
vide a suitable environment to enable this amazing ability in a closed containment:
the bioreactor.
Microorganisms form complex molecules, which can be produced in chemical
processes only with a great effort in many consecutive inefficient conversion steps.
 1.6 Characteristics of living cells – unique selling points from an engineering view | 15

Response to Metabolism
physical and complex, highly
chemical signals specific reacons

The cell

Self-organizaon Autocatalysis
formaon of spaal biomass produces itself
structures

Fig. 1.11: Unique selling points of microorganisms as catalysts: these include self-reproduction,
synthesis of complex molecules, autonomous formation of spatial structures and the response to
environmental signals.

Complexity in this context means sequential order of monomers in polymers, and

regiospecific and enantioselective conversions. For many pharmaceuticals no other
production routes exist apart from production by bioprocesses. This is indeed a very
strong unique selling point for their employment.
Cells are in no way ideally mixed microreactors. They maintain a highly complex
compartmentation and spatial order from organelles over enzyme complexes, facil-
itating ordered metabolite flux down to transport of molecules to distinct locations.
Such spatial structures are not only necessary to decouple the hundreds of enzymatic
steps but can also be employed directly for production purposes. Long chains of sep-
arated chemical reactors would be necessary to mimic this unique feature.
Cells react to physical and chemical signals by adaption of their physiology and
their metabolic pattern to the environment. Chemical signals include media concen-
trations leading to overshoot metabolism on the one hand or starvation on the other.
Specific chemicals induce specific reactions like excretion of specific products or are
employed as inhibitors. Physical signals include temperature, mechanical stress or
light. Such signals are also employed during bioprocesses e.g., to switch the cellular
metabolisms from growth to production mode.
For all of these unique selling points examples will be given in the following chap-
ters.
16 | 1 Introduction – a thread through this book

1.7 Exercises, questions and suggestions

1. Make a shopping tour and look at the list of ingredients on the cans, packets
and bottles. Try to find out which of the items are biotechnologically derived.
2. What are the ‘parts and models’ of living organisms in the sense of the OECD
definition?
3. Which unit operations in the yeast production process are located in the field
of mechanical, chemical or thermal process engineering?
4. Bioproduction takes place for many fields of application: food produc-
tion, agriculture, technical additives, environmental and energetic processes,
pharmaceuticals, commodity conversion, etc. Collect examples of processes
and products from this book or other resources.

1. Corn syrup (glucose, high fructose corn), aspartame, citric acid, glutamate, glu-
conate, yeast extract, xanthan, ectoine, PUFA (poly-unsaturated fatty acids), vi-
tamins B2 (riboflavin), B12, C, Q10, lysozyme, gluconic acid, etc.
2. Enzymes, mitochondria, ‘in silico models’, computer models, biosimilars, etc.
3. Mechanical: mixing, filtration; chemical: reaction technology; thermal: cooling,
drying.
4. Food: food thickener, flavorings, starter cultures; agriculture: feed proteins and
amino acids, crop protection products; additives: lubricants, biopolymers, wash-
ing powder, biocatalysts; environment and energy: wastewater, biogas, biofuels,
soil decontamination; pharmaceuticals: therapy and diagnostics, antibiotics,
hormones, vaccines, antibodies; commodities: degradation of lignocellulosic
materials, bioleaching.
2 Biosystems – microorganisms and other
biocatalysts
Omne vivum ex vivo (Latin: All life is from life). Until the nineteenth century this in-
sight was a subject of discussion until Louis Pasteur and other researchers brought
it to a head. Nowadays we accept this law of biogenesis as a matter of course. Strain
development, including directed evolution and genetic/metabolic engineering, is an
integrated step in upstream process development.
Usually process engineers get production strains from the upstream department.
Nevertheless, communication between biologists, biotechnologists and bioengineers
is necessary to consider the strengths, weaknesses, opportunities and threats of dif-
ferent production scenarios, including strain selection, in the spirit of integrating the
different sciences. This early decision has far reaching consequences on the bioreac-
tion and downstream steps, and ultimately the commercial success. According to the
definition of biotechnology, biosystems include microorganisms and plant and animal
cells, but also parts of these. Single cells in suspension culture do not represent the
only form in which organisms are used in bioprocesses; different kinds of structures
are also employed. An overview is given in Figure 2.1 and shortly discussed below.
It is not the aim of this chapter to give a comprehensive treatise of the biology of mi-
croorganisms, but to give ideas about the diversity of possible production systems and
to sensitize the reader to opportunities, risks and technical constraints in bioprocess
design.

Enzymes
Parts
Organelles

Bacteria
Prokaryotes
Archaea

Yeasts

Biosystems Filam. Fungi

Eukaryotes Plant Cells

Microalgae

Animal Cells Fig. 2.1: Overview of different biosystems including

Biofilms prokaryotic and eukaryotic microorganisms, plant
Consora and animal cells, parts of them like enzymes and
Tissues structures of them like biofilms.

https://doi.org/10.1515/9783110315394-002
18 | 2 Biosystems – microorganisms and other biocatalysts

2.1 Motivation – a first bioengineering view of microbiology

Below we briefly summarize aspects of process oriented strain selection for forward-
looking technical planning.
Agricultural crops and domestic animals have been farmed and bred for centuries
and are adapted to the artificial environment provided by people. This concept of do-
mestication can also be transferred to microorganisms. Some of them are employed in
traditional bioprocesses or have been used in modern biotechnology for decades. For
these strains, well elaborated cultivation procedures are established, and process de-
sign follows standard rules. Undesired side effects like byproduct formation, sticking
to reactor walls, flocculation, or strong foaming are reduced by long term selection.
These are arguments to retain such strains as production platforms for different prod-
ucts. For newly screened strains, their behavior in the reactor has to be tested in order
to select the best strains and prevent disappointing results during fermentation and
cell harvesting.
Process safety is a major concern in biotechnology and is influenced by the choice
of the production organism. The simplest case is microorganisms generally regarded
as safe (GRAS status). Cultivation of genetically engineered organisms requires spe-
cific measures for the production plant, the reactors, but also for handling and docu-
mentation. This holds especially for pharmaceutical products. Potentially pathogenic
organisms also demand exacting safety regulations. The decision is made for each
strain; even some E. coli strains are classified as potentially pathogen. Consequently,
process engineering is facing many complicated detailed problems and cost issues.
In connection with genetically engineered organisms, process stability also has
to be investigated. For example, loss of plasmids can slowly reduce productivity in
the long term during continuous cultivations. But wild strains can also be subjected
to spontaneous mutations or at least phenotypical changes. Microorganisms running
through a life cycle like sporulation need special attention in process control.
Size, cell wall and cell shape determine mechanical stability of cells in the ag-
itated reactor. Turbulence causes the formation of eddies of many different length
scales (Kolmogorov length scale), where the smallest structures are in the range of
50 µm in bioreactors. Bacteria and yeasts being much smaller are more or less on the
safe side. Furthermore, efficiency of downstream steps especially cell separation and
cell disruption depend on cell size. Filtration and centrifugation becomes more and
more difficult for biomass with decreasing cell size.
Strain selection determines medium composition. Consequently, it includes tests
for growth on technically cheap media but also for consumption of most of the car-
bon sources present in the medium. This holds especially for low value products. Re-
sistance against osmotic pressure or occasionally occurring inhibiting chemicals in
the medium are further demands on microorganisms used for production. Tolerance
against low pH values can be an issue and helps suppress contaminations.
 2.2 Diversity of biosystems – appearance in technical environments | 19

2.2 Diversity of biosystems – appearance in technical

environments

Microorganisms appear in an incredible diversity in nature. Even a short overview is

practically impossible in the framework of this book. The following paragraphs list
some technically important groups and emphasize the different aspects for bioprocess
engineering.

2.2.1 Enzymes – the universal toolbox for biocatalysis

Biocatalysis is the chemical process through which biological catalysts, which can
be one or more enzymes (or even cells), perform chemical reactions between organic
components. Thereby the catalyst can significantly lower the activation energy of a re-
action, so that the speed of the reaction is increased, whereby the catalyst itself is not
used up. Generally, protein based enzymes are responsible for catalytic reactions in
all living organisms and are considered the common catalytic units that form the basis
for biotechnological transformations processes, since they can increase the reaction
speed by up to a factor of 108 to 1020 . Classical fields of application can be found in
food and drink processing, where the production of wine, beer, cheese, etc. is depen-
dent on the effects of microorganisms or enzymes. In fact, biocatalysis underpins some
of the oldest chemical transformations known to humans, with the oldest records of
brewing used by Sumerians about 6000 years ago. The Incas used salvia (probably
without a deeper understanding of the mechanisms acting behind) as a source of the
enzyme amylase, used to break down starch from corn for the production of a sort of
beer called chicha. The identification of enzymes as the catalyst of a specific reaction,
however, was first successfully described in 1877, when German physiologist Wilhelm
Kühne defined the term enzyme, which was later used to describe the chemical activity
of nonliving substances. Since then a lot of efforts have been made to gain knowledge
of the working mechanisms of the enzymatic catalysis, and therefore enzymes also
play a key role in modern biotechnological applications (e.g., cheese thickening was
one of the first processes using recombinant enzymes) Figure 2.2.
Protein/enzyme mass is usually given by Dalton (Da) or kilo Dalton (kDa), and is
typically in the range of 3 to more than 1000 kDa. The unit Da corresponds to atom
mass u (u = 1/12 of 12 C carbon atom mass), and can be used for calculations on a
mol basis (1 mol 12 C = 12 g). Another important factor for enzyme applications is the
enzymatic activity. Usually the activity of enzymes is given in units (U) and stands for
the amount that can be catalyzed per µmol substrate within one minute. For example,
if hexokinase (first step of glycolysis) activity is given by 350 U/mg enzyme, it means
that 1 mg of the enzyme (hexokinase) will catalyze the phosphorylation of 350 µmol
substrate (glucose) within one min under optimal conditions (which are normally also
stated by the supplier). Another way to express maximal possible enzyme activity is
20 | 2 Biosystems – microorganisms and other biocatalysts

Fig. 2.2: Structural visualization of cellulase enzyme from bacteria Thermomonospora fusca
(TfCelE4). (© Olga Blifernez-Klassen). (a) Ribbon structure model with artificially colorized α-helices
and β-sheets shown. (b) A surface plot model where some functional groups are colorized: substrate
binding domain (gray), catalytic domain (dark gray), and conserved amino acids within the catalytic
domain (red, yellow).

the turnover number (also termed k cat ), which specifies the maximum number of cat-
alytic substrate conversions per enzyme in one second.
Enzymes often require cofactors to function, and some cofactors like metal
ions (Cu, Fe, Mg,) do not change during the reaction. Other cofactors like ATP or
NADH(P)+ +H+ are used up by the catalytic reaction, and can be designated as cosub-
strates. These cosubstrates have to be regenerated with another coupled enzymatic
system. This needs to catalyze an exothermic reaction, from which the substrate is
lost. Ongoing research is developing such processes robust enough for industrial
application.
A commonly accepted top level classification based on the reaction mechanism
lists oxidoreductases (oxidation/reduction), transferases (transfer a functional group
e.g., phosphate), hydrolases (hydrolysis of various bonds), lyases (cleave various
bonds other than hydrolysis and oxidation), isomerases (catalyze isomerization), and
ligases (join two molecules with covalent bonds). A selection of commercially applied
enzymes is given in Table 2.1 covering different reaction groups and application fields.
Features for favorable application of enzymes can be identified from this list. First
of all, the ability to distinguish between similar substrate molecules is one of the most
important advantages of a biocatalyst. This selectivity is often functional group spe-
cific (chemoselective) or the enzymes may distinguish between functional groups that
are chemically situated in different regions of the substrate molecule (regioselective or
stereospecific). This high selectivity may offer several benefits to the overall process,
such as high catalytic efficiencies and mild operational conditions, minimized side
reactions, easier separation, and fewer environmental problems. However, this also
 2.2 Diversity of biosystems – appearance in technical environments | 21

Tab. 2.1: Selected list of common technical enzymes.

Scientific name Function Application Comments

Protease Protein Washing powder Bacillus spec.
degradation
Amylase Starch Food industry, bioethanol E. coli, B. subtilis,
degradation industry, textile industry, B. amyloliquefaciens,
bakery industry B. licheniformis
Chymosin Proteolytic Cheese thickening, Aspergillus sp., E. coli
substitution of rennin
Phytase Hydrolysis of Additive in animal food Bacillus spec.,
phytic acids Lactobacillus plantarum
Cellulase Cellulose Laundry detergent, pulp Clostridium sp., Cellulomonas sp.
degradation and paper Industry

covers a unique selling point that makes chemical alternatives with unspecific cata-
lysts difficult if not impossible. Applications in the food industry take advantage of
the fact that enzymes can be left inside the product. In fact, it would be impossible to
remove them in many cases like cheese. For fluid and molecular disperse systems dif-
ferent methods have been developed for enzyme immobilization. This avoids mixing
into the product and enzyme losses during the process. Due to their relatively simple
applicability and low costs, the employment of enzymes outside classical biotechno-
logical production is an emerging field.

2.2.2 Bacteria – organisms for all seasons

Bacteria are present ubiquitously in the environment and are key players in global or-
ganic and inorganic material cycles. Traditionally they are involved in food fermenta-
tion, including different dairy products. On the other hand, they are part of the micro-
biota and reside on our skin and inside the intestine. Humans have been suffering from
bacterial diseases for millennia and are still suffering. Therefore, bacteria are targeted
by biotechnological research with respect to finding antibiotic activities. Antonie van
Leeuwenhoek (1632–1723) was the first to watch bacteria in his strongly improved mi-
croscope in 1674. This year is thus regarded as the beginning of microbiology. It is
noteworthy to mention that Leeuwenhoek was not only an enthusiastic scientist but
also a good communicator and made society aware of his findings. A modern electron
microscopy image Figure 2.3 shows the typical appearance of bacterial cells.
Besides their phylogenetic classification, bacteria species can be divided into two
groups regarding the chemical and physical properties of their cell walls, measurable
by so-called Gram staining. Both bacteria groups possess rigid polysaccharide (pep-
tidoglycan, also known as murein) layer within the cell wall in order to compensate
22 | 2 Biosystems – microorganisms and other biocatalysts

Fig. 2.3: Bacteria cell appearance under scanning electron microscope (SEM). (a) A cluster of JCVI-
syn3.0 cells, artificial cells with minimal bacterial genome (Craig Venter, © Science) showing spher-
ical structures of varying sizes. (b) Group of gram-negative Escherichia coli bacteria (here strain
O157:H7), artificially colorized for better visualization (© CDC).

for hydrostatic pressure (~ 2 bars). The peptidoglycan layer is more pronounced in

gram-positive bacteria, since gram-negative bacteria possess an additional (to the cy-
toplasm membrane) outer membrane, on the top of a thin peptidoglycan layer. These
differences in the cell wall structure are responsible for positive or negative staining
with crystal violet dye, commonly used for discrimination of gram-positive and gram-
negative bacteria.
The size of technically relevant bacteria is about 1 µm diameter × 2 µm length mak-
ing a volume of about 1 μm3 . These dimensions make filtration and centrifugation dif-
ficult without pretreatment of the suspension.
The huge diversity of metabolic patterns is consequently represented in technical
processes and products. A list covering some technically relevant strains and fields of
application are listed in Table 2.2.
The largest product examples with respect to quantity are small metabolites.
Anaerobic processes are operated for solvent (acetone/butanol) production as well as
for bulk products for the chemical industry. A prominent example is 1,3-propandiole,
which is used as a precursor for polymerization. The organic acid lactate is also an
important example not only for the food industry but as recently discussed for poly-
lactate production as bioplastics. Aerobic production of organic acids is employed for
amino acids, where glutamate is an ingredient in many consumer foods. Bacillus is
a natural producer of extracellular proteins making it a preferred candidate for tech-
nical enzymes such as hydrolases like protease, amylase, or cellulase. Recombinant
proteins are formed by E. coli and intracellularly stored as inclusion bodies, the most
prominent example being insulin. The production of antibody fragments is also an im-
portant application. Bacteria can target proteins into the cytoplasm, the periplasm, or
the extracellular environment. This has decisive consequences for product recovery.
Other applications of bacterial metabolic performance cover wastewater treatment.
 2.2 Diversity of biosystems – appearance in technical environments | 23

Tab. 2.2: Selected list of bacteria of industrial importance.

Scientific name Biology Application Comments

Escherichia coli Intestinal Recombinant Insulin production since 1982,
proteins model for genetic engineering
Corynebacterium Soil bacterium Amino acids Since 1957 in Japan as a glutamate
glutamicum producer, model for metabolic
engineering
Bacillus subtilis Soil bacterium Extracellular Proteases, lipases, etc. for washing
technical enzymes powder
Lactobacillus Milk products and Lactic acids, Starter cultures in food industry
intestinal dairy products
Streptomyces Filamentous soil Streptomycin Many industrially produced
bacterium antibiotic agents, streptomycin
(1952), ivermectin (against worm
infections, 2015)

2.2.3 Archaea – the underestimated extremists

Archaea are prokaryotes with a cell size ranging from 0.1 to 15 µm in diameter and
do not contain a cell nucleus or any other membrane-bound organelles. Despite their
general morphological similarity to bacteria, they represent a fundamentally differ-
ent biochemical and phylogenetic group compared to the bacteria kingdom, and pos-
sess genes and several metabolic pathways that are more closely related to those of
eukaryotes. Cell walls of archaea do not contain peptidoglycan, and are instead com-
posed of different polysaccharides, proteins and glycoproteins. Some methanogenic
archaea contain a polysaccharide (pseudomurein) that is similar to the peptidogly-
can of the bacteria. This difference in the structure is sufficient to prevent cell wall
degradation by lysozyme, which is efficiently used for bacteriolysis. However, most
common archaea are enclosed only by a so-called S-layer (consisting only of proteins
and glycoproteins), which seems to be sufficient to prevent osmotic lysis of the cells.
The name archaea means ‘ancient things’ and points towards extreme environments
where these organisms can often be found (temp. > 80 °C, pH of 0.7, high salt concen-
trations and anaerobic conditions), which are similar to conditions on the early Earth.
The ability to proliferate and be metabolically active under various extreme conditions
makes these organisms very interesting for biotechnological applications. Overexpres-
sion of archaeal enzymes produced e.g., by Pyrococcus furiosus, which are resistant ei-
ther to heat and/or acidity and/or alkalinity in established mesophilic overproducing
hosts (bacteria, yeast), have found many applications in modern biotechnology. The
thermostable DNA polymerase (the key enzyme in polymerase chain reaction [PCR])
gene is originally from hyperthermophilic archaea and is produced for commercial
purposes in E. coli. The industry uses recombinant enzymes (amylases and galactosi-
24 | 2 Biosystems – microorganisms and other biocatalysts

Tab. 2.3: Archaea strains used for industrial product formation.

Scientific name Biology Commercial products Comments

Haloferax Extreme halophilic exopolysaccharides (EPS) Can be cultured not only on
mediterane poly(3-hydroxy- glucose but also on extruded
butyrate-co-3-hydroxy- cornstarch, rice bran, or other
valerate) (PHBV) hydrolyzed organic wastes
Haloarcula sp. Extreme halophilic poly(3-hydroxy-butyrate) Can be cultured on
IRU1 (PHB) petrochemical wastewater,
high PHB yield

dases) from other Phyrococcus species for food processing at high temperatures, for
example during the production of low lactose milk.
Despite the versatile use of recombinant archaeal enzymes, the use of archaea
cultures seems to be less variable. Although only a few applications of living archaea
exist, they are all applied in larger dimensions. So, for example, the biological produc-
tion of methane (biomethane production) in various anaerobic processes is performed
by archaea. Although the process of anaerobic digestion is mainly driven by a versa-
tile microbial community of facultative and obligate anaerobe bacteria, the final step
(methanogenesis, Section 9.4 on biogas) is performed exclusively by archaea (thus
far). Other thermophile archaea (e.g., Sulfolobus metallicus or Metallosphaera sedula)
are used in industrial bioleaching processes, where metals are extracted from the ores
(Section 9.5 on bioleaching). Furthermore, industrial production of PHB, PHBV and
different EPS (extracellular polymeric substances) can be performed with extreme
halophile archaea species (Table 2.3).

2.2.4 Yeast – the workhorse of bioprocesses

Baker’s yeast is probably the microorganism most closely related to human culture,
serving as a leavening agent in breadmaking and providing fermentation activity for
brewing and winemaking. Yeasts grow at quite low pH values of 4–5, suppressing most
of the potentially competing bacteria. This ‘self-sterilization’ property may be one rea-
son why humans could manage dough maturing without specific precautions. Ethanol
formation during brewing also supports self-sterilization.
In contrast to bacteria, yeasts do not propagate by cell fission but by budding, an
asymmetric division process. During unequal budding a ‘daughter cell’ is formed on
the ‘parent cell’ and grows. It separates from the parent cell before it reaches the same
size. In the meantime new buds can be formed, leading to small primary aggregates
(flocs). Light- and electron-microscopic pictures of an unequally budding yeast are
shown in Figure 2.4.
 2.2 Diversity of biosystems – appearance in technical environments | 25

Fig. 2.4: Examples of yeast cell appearance. (a) Yeast cells (Pichia pastoris) cell appearance under
light microscope (© Viktor Klassen). (b) Culture of yeast cells (Saccharomyces cerevisiae) depicted
by SEM. Both are proliferating by unequal budding (©Ramasamy Patchamuthu).

With the size of about 5 µm diameter, the flocs being accordingly bigger, yeast cul-
tures are assessable by centrifugation on a technical scale. This is a good example
of a process where strain selection and growth conditions have a direct influence on
downstream processing.
This process is of technical meaning as parent cells exhibit a higher fermenta-
tive activity and stability in the dough, while the buds grow potentially faster during
production. Filtration after yeast production and sedimentation or floatation during
fermentation is largely influenced by the size of the flocs. Yeast can be regarded as mi-
croorganisms with a high degree of ‘domestication’. For 200 years brewers and bakers
conducted more or less intuitive strain selection to bring up process relevant features,
characterizing today’s production strains. A segregated view on the cell population
helps to understand yeast culture. This concerns not only parent and daughter cells
but also cell size. Mother cells show different numbers of bud scars, so we can speak
in a generalized sense about cell age, in contrast to the ‘immortal’ bacteria.
But yeast have more abilities than making ethanol and carbon dioxide. Some
other strains and their respective applications in modern bioprocess engineering are
given in Table 2.4.
Yeasts are subject to one of the earliest regulations in history. In the so-called
‘Reinheitsgebot’ (German Beer Purity Law, 1516 Bavarian law based on earlier regu-
lation) it was stated: “Furthermore, we wish to emphasize that in future in all cities,
markets and in the country, the only ingredients used for the brewing of beer must be
barley, hops and water.” Note that yeast is not mentioned here as the biological prin-
ciple of fermentation was not known at that time. Yeast came into beer from the air
or from impure fermentation vessels. Until now, the integration of bioprocesses into
society has meant to obey or to propose regulations, which influences process design.
Yeast production will be further outlined in this book as a running example.
26 | 2 Biosystems – microorganisms and other biocatalysts

Tab. 2.4: Industrial applications of common yeast stains.

Scientific name Biology Commercial products Comments

Saccharomyces Unequal bud- Yeast for baking and brewing Used for millennia
cerevisiae ding, aerobic and for alcoholic bever-
anaerobic ages
Pichia pastoris Methylotroph, Recombinant proteins for Strong natural
(Komagataella methanol for pharmaceutical/food industry promotor
phaffii) induction
Ogataea Methylotroph, Enzymes (phytase, hexose oxidase, Tightly regulated
angusta thermotolerant lipase) as food/beverage additive native promoters
(Hansenula (50 °C), Proteins/peptides (hepatitis B
polymorpha) surface antigen, hirudin,
insulin-like growth factor for
biopharmaceutical use
Candida Single cell protein (SCP) SCP also Pichia

2.2.5 Filamentous fungi – the broad-spectrum chemists

Filamentous fungi employed in bioprocesses do not grow as suspended single cells,

but in long threads consisting of a chain of cells. These hyphae grow only at the tips
(apices), while the older cells do not divide any more. However, they do contribute to
metabolism and transport the products to the growing tip. Depending on growth con-
ditions and length of a hypha, side branches can be generated, thus forming a braid
called a mycelium. While the hyphae have a typical diameter of 5 µm, the whole or-
ganism can reach macroscopic dimensions. Under natural conditions in the soil the
mycelium can cover several square meters. The Humongous Fungus (Armillaria so-
lidipes) is known to be one of the largest living organisms. In the Malheur National
Forest scientists have found a single specimen covering 8.4 km2 (3.4 square miles). It
grows underground, digesting dead wood but also attacking living trees. The weight
is estimated to be 660 t and the age 2400 years. That it is indeed a single organism is
proven by DNA sequencing.
In bioreactors, the growth habit looks different Figure 2.5. Growing tips experience
strong shear forces, bending them back to other threads or even letting them break.
From the fragments new hyphae can grow. Finally, a hairy ball of wool – the pellet –
emerges, which can be several mm in diameter. This has severe consequences for the
process. Oxygen is taken up by the outer layers of the pellet and cannot diffuse fast
enough into the center. This leads to oxygen depletion and death of the affected cells.
Usually, pellets are hollow. Filtration on the other hand is easier, as particles are large
and the filaments quite rigid (Table 2.5).
The discovery of penicillin by Alexander Fleming (1881–1955) is well known and
features in many history films. It is a prime example of the integration of science into
 2.2 Diversity of biosystems – appearance in technical environments | 27

Fig. 2.5: Appearance of filamentous fungi. (a) Mycelial biomass pellets of Ganoderma lucidum are of
7–12 mm in diameter after submerged fermentation (© Marian Petre). (b) Scanning electron micro-
scope image of a fruiting body of the fungus Aspergillus niger, small spheres on the surface of the
fruiting body are spores (reproductive cells) (©Mogana).

Tab. 2.5: The two most technically important filamentous fungi.

Scientific name Biology Commercial products Comments

Aspergillus Filamentous Glucoamylase and other enzymes, Generally recognized
niger Fungus, also citric acid, as safe (GRAS)
known as black gluconic acid Genome fully
mold sequenced (2007)
Penicillium Filamentous β-lactam antibiotics (e.g., Genome fully
chrysogenum fungus penicillin) sequenced (2008)

society. After the first observations and isolation of Penicillin in 1928, it took ten years
until the world took notice of this great discovery. Process development started in 1939
and penicillin was produced in a large scale, being available to soldiers in 1944. The
impetus was the urgent need for a wound healing therapy during the Second World
War. Today’s discussion is still about “technology push versus market pull” or “fast
applied research versus basic research.”

2.2.6 Microalgae – the solar cell factory

Microalgae are unicellular plants ordered in several completely different morphologi-

cal genera. The term microalgae does not describe a well-defined biological group, but
more a habit; in a wider sense it also includes photosynthetic cyanobacteria. A picture
of the prokaryotic microalgae Spirulina is shown in Figure 2.6 (a). Microalgae account
for 50% of the global oxygen production. Furthermore, they are the basis of the aquatic
food chain. Polyunsaturated fatty acids derived from fish for our daily diet originate
28 | 2 Biosystems – microorganisms and other biocatalysts

Fig. 2.6: Different appearance of microalgae. (a) Arthrospira platensis (Spirulina) apearence under
light microscope (© Viktor Klassen). (b) An algae (Emiliania huxleyi) bloom off the southern coast of
Devon and Cornwall in England, satellite image (©Steve Groom).

from microalgae. The same holds for the red color of crustaceans and astonishingly of
flamingos, originating from microalgal carotenoids.
From an engineering view, microalgae grow in suspension culture when supplied
with light for photosynthesis. The unique selling point is of course the use of sunlight
as an unlimited available energy source and carbon dioxide as a carbon source. Un-
fortunately, technical realization is not so easy. Microalgae biomass is until now pro-
duced mainly in open ponds located in sunny regions. The annual production of food
supplements with traditional strains (Table 2.6) amounts to several thousand tons
only. Fine chemicals like carotenoids are used for pharmaceuticals and cosmetics. It is
another specialty of the algal cell to accumulate large quantities of these strongly re-
ducing compounds. Even if the genes were expressed in bacteria, a high product titer
could not be obtained. High value compounds have even been produced recently with
artificial illumination by light-emitting diodes (LEDs). However, there are still huge ex-
pectations in the direction of producing large amounts of biomass for feed, food and
biofuels to serve an increasing world population.

Tab. 2.6: The most commonly known microalgae; more examples are given in the case study on
microalgae.

Scientific name Biology Commercial products Comments

Chlorella Eukaryotic, Food supplement Traditionally approved for food

vulgaris green alga
Arthrospira Prokaryotic Food supplement Commonly called ‘Spirulina’, traditionally
platensis approved for food
 2.2 Diversity of biosystems – appearance in technical environments | 29

In fact, microalgae seem to be the only realistic alternative to produce renewable

biomass without competition with existing terrestrial crop plants. Production in
closed photobioreactors produces more than five times more biomass per hectare area.
One of the reasons is that the unicellular organisms do not form roots and stems. Culti-
vation can be done in dry areas due to reduced water evaporation and to use of saltwa-
ter. Ongoing research is coming closer and closer to the goal of economically feasible
mass production of microalgae. Application in fish feed mimics the food chain, thus
reducing the need for fishmeal. Proteins and vitamins from microalgae are a chance
for developing countries to reduce malnutrition. Finally, solar biofuels may substitute
fossil fuels. More about this exciting new developments is outlined in the case study.

2.2.7 Plant cells – slow but resourceful

Since the Neolithic period people have used plants to cure diseases. From corpse dis-
coveries it is known that moss cushions were carried during travelling, probably to put
on wounds for healing because of their (of course not recognized) antimicrobial activ-
ity. The use of salicin from the bark of the willow tree (Salix sp.) against headache has
been documented since the Middle Ages; today it is chemically produced in a mod-
ified form as acetylsalicylic acid. The indigenous population in North America used
purple coneflower Echinacea against inflammation of sores; nowadays extracts are
commercially available in each drugstore as immune stimulants. Direct cultivation of
the respective plants is subject to seasonal changes in availability and drug content.
Furthermore, it is difficult to get approval to use complex natural mixtures of active
compounds as drugs. Instead, pharmaceutical companies isolate active substances
and produce them chemically or use them as lead structures in drug discovery.
Nevertheless, there is ongoing research to employ plant cell cultures for produc-
tion as callus (unorganized proliferative mass of cells), meristem (from apical shoot)
culture, or as hairy root (induced by Agrobacterium) in mostly heterotroph suspension
or surface culture Figure 2.7.
Some examples with current technical application or systems under research are
given in Table 2.7.
According to their natural function, plants cells produce bioactive secondary
metabolites. Although 50,000 plant species worldwide are used for medical appli-
cations, only very few plant cell applications have reached commercial scale. Some
companies use special reactors where the tissues are sprinkled with medium to pro-
duce a variety of secondary products for pharmaceuticals and cosmetics. The two
first examples in Table 2.7 represent the main applications. The reason for the limited
distribution is the high cost due to low productivities. In the case of Taxus, agricul-
tural production as the competing approach is obviously not feasible, which seems to
be the case for Lithospermum as well. In other cases the bioactive compound can be
synthetized chemically.
30 | 2 Biosystems – microorganisms and other biocatalysts

50 μm 4 cm

(a) (b)

Fig. 2.7: Examples of plant cells/tissues. (a) Cells in a filament of Echinacea purpurea in heterotroph
suspension culture. The biggest part of the cell is filled with the vacuole. The kernel is connected
to the cell wall via by plasma strings. (b) Hairy roots of Beta vulgaris (beet) grown in a petri dish
(© E. Steingröver).

Tab. 2.7: Important plant cell based productions systems.

Scientific name Biology Commercial products Comments

Taxus brevifolia Plant cell from Paclitaxel anticarcino- Production since 1993 in Ger-
Pacific yew tree genic many, up to 75,000 l scale,
1 bn €/$ market volume (2016)
Lithospermum Purple gromwell Shikonin, antibiotic Since 1983 in Japan (purpura,
erythrorhizon properties (external red circle in flag)
ulcers) red color for
cosmetics and textiles
Physcomitrella Spreading earth Human identical hor- Research and diagnostics, first
patens moss, phototroph- mones clinical phase
ic cultivation

Phototrophic cultivation can claim some advantages for the production of recombi-
nant proteins as in the case of Physcomitrella (‘pharming’ = production of recombi-
nant proteins by farming of genetically engineered terrestrial plants). Cultivation is
based on pure mineral medium (with plant hormones), which makes cell separation
easy. The absence of organic carbon sources prevents contamination. Another intrin-
sic product safety aspect (compared to mammalian cell culture) is that no viruses are
known to affect both plants and humans. Products from plant cell culture other than
proteins are limited to compounds, which are difficult to express in other cell systems.
Plants have abilities for posttranslational modifications and can also excrete large pro-
teins. Recent developments aim for the production of human identical proteins with
respect to glycosylation and could therefore lead to a production platform as an alter-
native to animal cell culture.
 2.2 Diversity of biosystems – appearance in technical environments | 31

2.2.8 Animal cells – masters of protein decoration

Cell culture is the process by which cells are grown under controlled conditions, gen-
erally outside of their natural environment. In practice, the term ‘cell culture’ refers to
the culturing of cells derived from multicellular eukaryotes, especially animal cells, in
contrast to microbial cultures. In recent years, there has been an increase in the use of
mammalian cells as expression systems for the production of biopharmaceuticals like
antibodies, vaccines, hormones, and nucleic acids. The advantages are higher qual-
ity and efficiency in the production of human or humanized glycoproteins compared
with nonmammalian cells (Table 2.8).

50 μm 15 μm

(a) (b)

Fig. 2.8: Examples of animal cells visualized by immunofluorescence. (a) Hep G2 cells, cultured in a
tridimensional microcavity of 3D -KITChip (cytoceratin 18 [green], cell nuclei [blue]) (© E. Gottwald).
(b) Human embryonic kidney (HEK) 293 cells (F-actin [green], 11–15 µm nuclei [red], phospho-histone
H3 [yellow]) (©creativecommons).

Tab. 2.8: Animal cell based productions systems.

Scientific name Biology Commercial products Comments

Animal cell line Epithelial cell line de- Therapeutic Most important system for
Chinese ham- rived from the Chinese antibodies humanized recombinant
ster ovary (CHO) hamster ovary proteins
Animal cell line Fusion of white blood Monoclonal Permanent (immortal) cell
Hybridoma cells (B-cells) and antibodies (mAbs) lines due to origin from
myeloma cells cancer cell
Insect cell line Isolate of ovarian tis- Various recombinant Baculovirus vectors are used
podoptera sue of fall armyworm proteins as a transgen shuttle system
frugiperda Sf9
32 | 2 Biosystems – microorganisms and other biocatalysts

Mammalian cells, with a typical diameter of 15 µm Figure 2.8 (b), are quite large com-
pared to yeast and bacteria. They grow quite slowly (t d about 10–20 h) and require a
complex and expensive growth medium. Beside sugars, they contain amino acids and
animal serum prepared from blood. Mammalian cells possess a cell membrane but no
rigid cell wall. This means that together with their large size that the cells are very sen-
sitive to shear stress and cannot be cultivated in continuously stirred tank reactors. In
many cases bubble free aeration has to be provided, see Section 5.5 on aeration. For
production the cells can be propagated in suspension culture, but for high yield cul-
ture are often attached as monolayers (2D) on the surface of small spheres (⌀200 μm)
used as microcarriers. As a technical limit, up to 200 million cells per ml can be ob-
tained, and reactor sizes up to 6 m3 are in use. For medical research, cells are grown
more and more in so-called 3D culture e.g., in cavities of silicon chips Figure 2.8 (a).

2.2.9 Organized communities – making rich booty as a consortium

Not all bioprocesses work with monoseptic cultures containing only one species.
In fact, in nature that is the great exception. In biogas plants carbohydrates are
hydrolyzed to sugars, degraded to carbonic acids, and then further metabolized to
hydrogen and methane. Each of these steps is carried out by several different mi-
croorganisms specialized to a few specific reactions. In this way the different strains
support each other: the community is much more efficient than only one strain. It has
turned out to be a problem when cutting off one branch of the degradation pattern
e.g., to get pure hydrogen. Such communities have for the time being not been fully
understood, despite modern molecular methods being applied Figure 2.9.

100 μm

(a) (b)

Fig. 2.9: Visualization of biofilm structures. (a) 3D imaging of confocal laser microscopic characteri-
zation data presented as isosurface projection. Color allocation: red – nucleic acids, green – lectin-
specific EPS glycoconjugates (© M. Wagner). (b) Simplified diagram with three stages during biofilm
formation on a surface: attachment by ambient molecules (1), proliferation/EPS synthesis (2), senes-
cence/detachment (3) (©Uschi Obst).
 2.3 Cultivation conditions – environmental parameters to care about | 33

While ideally mixed cell suspensions are usually applied in bioprocesses, in nature
many microorganisms live attached to surfaces. The cells cover themselves with an
extracellular matrix consisting of self-produced extracellular polymeric substances
(EPSs), consisting mainly of polysaccharides and proteins as well as nucleic acids
and lipopolysaccharides in minor concentrations. This structured community forms
the biofilm. The EPSs protect the cells from drying out or from toxic substances. Even
in environments with extremely low substrate concentration, like in drinking water
tubes, the microorganisms can survive without being washed out. Inside a biofilm,
much higher volumetric productivities can sometimes be observed than in a bioreac-
tor. Biofilms are much more than an ensemble of cells: a lot of chemical communica-
tion is going on to organize the structure and cooperation between the different organ-
isms. This encouraged researchers to employ biofilms in technical systems. While in
wastewater it happens more or less naturally, artificially supported biofilms for pro-
duction are also under investigation. A substrate can be provided on the biofilm side in
direct contact or by diffusion through a membrane on which the film is attached. Cell
retention in continuous cultivations is one specific advantage, making a separation
unit superfluous.

2.3 Cultivation conditions – environmental parameters to care

about

Across all biological groups microorganisms make high demands on chemical and
physical parameters in their environment. Besides chemicals there are three other
environmental factors affecting microbial growth: temperature, pH value and water
availability. Appropriate environmental conditions have to be guaranteed in in the
bioreactor. Shifts in the environmental conditions can be an additional degree of free-
dom in process design to influence intentional growth and product formation. These
environmental factors are listed in the following paragraphs.

2.3.1 Primary nutritional groups

A commonly applied classification of primary nutritional groups is depicted in Fig-

ure 2.10. The two basic criteria are the energy source and the carbon source of the
organism. Energy can either be gathered from light energy (photo-) or from chemical
redox reactions (chemo-), based either on organic (organo-) or inorganic (litho-, Greek
lithos = stone) electron donors. The sources of carbon can be of organic (hetero-) or
inorganic (auto-, CO2 ) origin as well. The electron acceptor is also decisive for the tech-
nical realization of a process. The use of oxygen is referred to as aerobic respiration.
Processes where intracellular intermediates are used by the cells as electron acceptors
are called anaerobic processes or fermentation; however this latter term is used quite
34 | 2 Biosystems – microorganisms and other biocatalysts

Photo Energy Redox C-source

Energy
Chemo
CO2
C-Source
Organo Chemolithoauto-

Litho Chemolitho-
Redox
Organo Chemo- Mixo-
Chemoorgano-
All organisms
-trophs
Photoauto-
Photo-
Photohetero-

Fig. 2.10: Primary nutritional groups; the colors underpin the different aspects.

nonspecifically in scientific communication. Last but not least the metabolic pattern
where the cell uses inorganic electron acceptors, e.g., nitrate, is called anaerobic res-
piration.
Aerobic processes with e.g., glucose as a carbon and energy source make active
aeration of the culture necessary, but lead to the most intensive bioprocesses with
respect to volumetric productivity of biomass and products. About 50% of the carbon
from glucose is allocated for biomass.
Anaerobic processes are usually quite simple and lead only to low biomass pro-
duction containing about 5% of the carbon of glucose consumed. Most of the carbon,
up to 50%, is directed into the reduced intermediate, which is typically the desired end
product. We meet such processes in ethanol fermentation or biogas production. Strict
anaerobic bacteria need removal of oxygen from the feed medium. Facultative anaer-
obes can at least remove the oxygen by themselves. Microaerobic conditions can also
be applied, where oxygen respiration covers the maintenance energy demand of the
cells, but product formation occurs via anaerobic metabolism. Another way of using
the capabilities of facultative anaerobes (e.g., E. coli) is to start a batch aerobically to
yield high biomass titers and then switch to anaerobic conditions to produce organic
acids.

2.3.2 Temperature

Temperature dependence of chemical reaction constants k Comp [mol/s] are usually ap-
proximated by the Arrhenius equation (2.1) as given by Svante Arrhenius in 1889.
−E a ,Comp
k Comp = AComp ⋅ e RT (2.1)
 2.3 Cultivation conditions – environmental parameters to care about | 35

The two parameters A [mol/s], being a prefactor, and E a [kJ/mol], the activation en-
ergy, have to be determined experimentally for different enzymes or microorganisms.
As a rule of thumb, increasing the temperature by 10 °C leads to an increase of the
reaction rate by a factor of two to four. This underlines the importance of a good tem-
perature control. Physical processes like diffusion exhibit much weaker temperature
dependencies (e.g., 20%/10 °C). In microorganisms, hundreds of chemical reactions
occur in parallel coupled by transport steps like diffusion. In principle, even each sin-
gle reaction step making up observable enzyme kinetics (see Chapter 4 on kinetics)
has different values for E a . Nevertheless, the Arrhenius equation works quite well to
describe temperature dependent growth rates but only for temperatures below a dis-
tinct value. Above this temperature optimum reactions rates of enzymes or growth
rates of microorganisms decrease. In biochemical systems, the general assumption
of the Arrhenius equations being temperature-dependent collision and assignment of
an activation energy to molecular species are not entirely true. Macromolecules show
increasing mobility of single residues, and enzymes may partially lose their spatial
structure, leading to decreased substrate affinity. On the cell level membrane fluidity
can change or regulation of single pathways for adaption can occur.
To summarize these effects, different mathematical correlations to describe tem-
perature response have been proposed. They relate mainly to a maximum growth rate
(see chapter Chapter 4 on kinetics for definition) assuming that all other influences
are in a nonlimiting and noninhibiting range. The most commonly used approach is
the assumption of a decay rate, for which the rate constant increases with temperature
following again an Arrhenius equation (2.2).
−E react
Areact ⋅ e RT
μmax (T) = −∆Gdecay
(2.2)
1 + Adecay ⋅ e RT

This approximation has a defined optimum Figure 2.11 and decays (∆G = free en-
thalpy) in the direction of decreasing and increasing temperatures. The four unknown
parameters have to be estimated from data. Even knowing that such a correlation is
not clear with respect to reversible/irreversible effects and timely aspects for duration
of temperature impact, it is practically applicable at least for moderate temperature
changes in a bioprocess. Other correlations claiming better intuitive access correlate
cardinal growth temperatures (Tmin , Topt , Tmax ) to a smooth curve. The assumption
of a temperature range outside of which no growth is possible may be more realistic
than the Arrhenius formula. An example is given in Equation (2.3).
T opt −T opt
Tmax − T T − Tmin Tmax −Topt
μ(T) = μ max ⋅ ( )⋅( ) (2.3)
Tmax − Topt Topt − Tmin

Besides the three cardinal temperature values, as a fourth parameter the maximum
specific growth has to be determined from data.
36 | 2 Biosystems – microorganisms and other biocatalysts

2.0
temperature dependent growth
1.8

specific growth rate [h-1] 1.6

1.4

1.2

1.0

0.8

0.6

0.4

0.2

0.0
10 20 30 40 50 60

(a) temperature [°C]

2.0

1.8 specific growth rate

1.6
specific growth rate [h-1]

1.4

1.2

1.0

0.8

0.6

0.4

0.2

0.0
3 4 5 6 7 8 9 10 11
(b) pH

Fig. 2.11: Growth curve for E. coli as function of primary growth factors. (a) Temperature impact ac-
cording Arrhenius formula (equ. (2.3), A = 6.4 ⋅ 1014 , B = 1.38 ⋅ 1048 , E react = 86.4, ∆Gdecay = 290).
The decline from the optimum to higher temperatures is usually steeper than the one towards lower
temperatures, so Tmax is closer to Topt than Tmin . (b) pH impact according equ (2.4) is assumed to be
symmetrical (pHopt = 7.0, σ pH = 1.5).

Because of the immense importance of temperature for the physiology of microorgan-

isms and for their technical application, microorganisms are classified according to
their optimum temperature as listed in Table 2.9.
The Arrhenius equation is specific for each given strain in this classification. Nev-
ertheless, optimum growth rates are usually lower for extremophiles. Consequently, it
is sensible that enzymes of extremophiles are often cloned into mesophiles for produc-
tion. In other cases, like in the food industry, respective strains are used directly. Note
that temperature control is not a technical problem as such but optimum temperature
 2.3 Cultivation conditions – environmental parameters to care about | 37

Tab. 2.9: Definition and occurrence of habitats with different temperature levels.

Temperature TOpt [∘ C]
Range −15–+15 10–30 20–45 45–80 80–113
Designation -philes Psychro-, cryo- Psychro- Meso- Thermo- Hyperthermo-
(Greek = affection) tolerant
Occurance Ice, meltwater Intestine Warm Hot springs, vol-
animals lakes canic lakes
Remarks ≤ 0°C, high Standard in Hot water No organics
salt content bioreactors pipes present
Application Food maturing Volatile Standard ‘Thermo- Thermus aquaticus
(examples) products production zymes’, Taq Polymerase
protease

for a specific strain/medium combination may not exactly be known. Temperature

shifts are, on the other hand, applied to induce changes in the microbial metabolism.
Microorganisms react to overheating by production of heat shock proteins. This is ap-
plied for production of recombinant proteins in E. coli, where a temperature shift in-
duces the respective promotor. Other cells can show changes in their lipid profiles or
show other potentially useful reactions.
An aspect useful for integration of downstream and bioreaction is to strip a prod-
uct like ethanol directly during cultivation, thus saving energy for distillation. In this
case it turns out that ethanol and temperature have a negative impact on the cell wall,
which makes the process unfeasible.

2.3.3 Acidity and basicity

Each species shows its optimum growth at a specific pH value. Deviation from this
optimum will lead to a remarkable decrease in growth velocity. Technically speaking
it is not a big problem to adjust the pH value in the fresh medium before it is used.
During cultivation pH shifts can be compensated by an acid or base. Nevertheless, in
many cases it is not reliable to do so, for example if an organic acid is the product
(Table 2.10).
As in the case of temperature it is not possible to simply find mechanistic kinetics
for the impact of pH on enzymes and microorganisms. The degree of dissociation of
residues (e.g., COOH− , NH+3 ) and its effect on surface charge would be only one item.
In the case of enzymes the isoelectric point (pI value) is the measurable expression of
pH influences on the surface charge. Specific action on the active site alters the affinity
constants. To have at least a preliminary anticipation of the impact of pH changes the
correlation 2 pHopt −pH
− 12 ( )
μ(pH) = μ max ⋅ e σpH
(2.4)
38 | 2 Biosystems – microorganisms and other biocatalysts

Tab. 2.10: Designation and occurrence of habitats with different pH values.

pH [-]
Range <4 6–7 >8
Designation, Acido- Neutro- Alkali-
-philes
Occurance Geothermal areas Soil, freshwater habi- Soda lakes (Na2 CO3 ), high Ca2+
(e.g., Yellowstone), tats and oceans groundwaters
metal mining sites
Remarks Suppression of con- High organism diversity Suppression of contaminants
taminants
Application Bioleaching Standard cultivations Enzymes, e.g., paper industry,
detergents

Tab. 2.11: Designation and occurrence of habitats with different salt concentrations.

Salt concentration as NaCl [mol/L]

Range 0–1 0.2–2.0 0.4–3.5 2.0–5.2
Designation, Nonhalo- Slight halo- Moderate halo- Extreme halo-
-philes
Occurance Technical Seawater Lake Qarun, Little Mani- Hypersaline lakes Dead Sea,
media tou Lake, Lake Urmia Lake Assal, Don Juan Pond
Remarks Normal Marine aqua- β-carotene production Habitats for halophile
case culture (algae) (Dunaliella salina) organisms

can be used. We see immediately in Figure 2.11 that the representation is formally equal
to a normal distribution (Gaussian) where pHopt is the optimum pH value and dpH the
standard deviation with dpH often in the range of 1.05–1.5 pH units for growth and en-
zyme activity. Interestingly, a symmetric approach is feasible. As in statistics, it sum-
marizes many different influences that are more or less unknown. Validity of such
models should not be overstressed, but could be in the range of +/ − 2dpH according
to published data sets. For process development more specific but unfortunately also
formal kinetics have to be found for each case where relevant (Table 2.11).

2.3.4 High concentrations

As with high salt concentrations, high concentrations of organic carbon sources also
cause high osmotic pressures. Osmophilic organisms can live in such environments
because they protect themselves by the synthesis of osmoprotectants, which are small
soluble molecules that increase intracellular osmotic pressure to prevent water loss of
the cells by osmosis. Mainly yeasts but also some bacteria can stand high osmotic pres-
 2.4 Bringing microorganisms to work – the concept of a ‘cell factory’ | 39

Tab. 2.12: Classification of microorganisms with respect to osmotic concentration.

Osmotic concentration [mol/kg]

Range 0–1.0 1.0–2.0 2.0–4.0 4.0–8.0 >8.0
Water activity 1–0.98 0.98–0.96 0.96–0.92 0.92–0.84 <0.84
a w [-]
Environment Normal Seawater Molasses Jam Dried fruit
medium
Organisms Most bacteria Most bacteria Osmotolerant, Zygosaccha- Saccha-
baker’s yeast, romyces romyces
Lactobacillus rouxii

sures. In technical terms this is quantified by water activity a W , the vapor pressure of
the solution divided by the standard partial vapor pressure of water. The osmotic con-
centration is the molar concentration of molecules contributing to osmotic pressure,
which are ions and all dissolved organic molecules (Table 2.12).
Approximation of the impact of water activity on growth rate can be done by the
Gaussian function in a similar way to pH. It is superfluous to note that the mentioned
environmental parameters affect the cells in mutual interactions. Some principles on
how to handle such cases will become clear in the case studies; others are given in the
additional references.

2.4 Bringing microorganisms to work – the concept of a

‘cell factory’

Bioprocesses consist of timely and spatially highly ordered steps of transport, bio-
chemical reactions, mechanical and thermal separation, product formulation, and de-
livery. Most of these steps require a special unit operation with a special equipment.
The process itself may be part of a factory, where a pattern of transport and conversion
steps also apply on a higher level. The means are of course different: transport is down
by conveyer belts or trucks, central control is the administration, and so on. Looking
downwards to the smaller units of a bioprocess – the cells – again this ordered struc-
ture of basic transport and conversion steps apply again with different means. This
has brought up the idea of the cell factory as shown in Figure 2.12.
This analogy can teach us several things. All steps have to be balanced in terms of
turnover. It does not help us to build up more assembly belts if not enough raw ma-
terial is provided. This has to be ensured along the supply chain from an external de-
liverer, through receiving controls, processing steps and finally, loading onto the belt.
The product has to be packed, targeted and mailed to the customer. For cell engineer-
ing it means that overexpression of a product has to be accompanied by balancing
40 | 2 Biosystems – microorganisms and other biocatalysts

Nucleus
Management &
Central Controller Golgi Apparatus
Delivery Store
Mitochondrium
Power Plant

Ribosomes
Transport Assembly Line
Purchasing
Department

Chloroplast Vesicle
Solar Collector & Chemical Reactor
Fuel Generaon Storage & Transport

Plasma
Parts & Tools

Fig. 2.12: Visualization of the idea of a ‘cell factory’, where the different organelles or sites in the
cell can roughly be matched to the unit operations and devices in a factory.

the whole cell with respect to all other production steps. The ‘management’ has to
be convinced that the product, e.g., a recombinant protein, is worth making without
corrupting the central control, which is necessary to balance the other intracellular
processing steps.
The role of bioprocess engineering lies mainly in the supply chain by maintain-
ing optimal conditions in the reactor, supplying the necessary parts and receiving the
product. However, these steps are not trivial. Building blocks, which may help the cell
do its job, have to be identified and provided in the medium: it has to be ensured that
the cell can take up these parts and use them properly. Product formation shall not
overcharge metabolism in general. Separating the product means isolating e.g., a pro-
tein from thousands of others. While during downstream processing this is a serious
problem, the cell can master it to perfection. Coordination between molecular and pro-
cess engineering can lead to a successful decision as to which unit operations are done
in the cell factory and which of them has to be done on the macroscopic process level.
A prominent example is targeting recombinant proteins outside the cell, which makes
macroscopic separation much easier. On behalf of the scientists involved in discover-
ing the molecular basis for intracellular protein targeting, the name of Guenther Blobel
may be recalled here at the Nobel Prize laureate (1999) for the discovery that “proteins
have intrinsic signals that govern their transport and localization in the cell.”
3 Media – supplying microorganisms with a
comfortable environment and building blocks
for growth
As well as the choice of the production strain, the medium has to be provided. The
spectrum ranges from synthetically composed media to complex natural media de-
pending on purpose and scale of cultivation. Media design is based on biological infor-
mation regarding stoichiometry and kinetics and can be straightforward using respec-
tive calculation. This approach goes only to a certain point, from after which strain
specific information and data based on experiences have to be respected. For further
optimization modern biotechnological techniques can be applied. In this chapter ba-
sic ideas for media design are given and specified by calculations and data.

3.1 Media – the material basis of the process

The main aspects of media design are listed in this paragraph. Specific examples will
be given in the case studies. A culture medium is by definition a chemical matrix of
substances designed to support growth of microorganisms. Minimal media or syn-
thetic media are based on known chemicals and contain only necessary components.
Complex media can contain many auxiliary materials to support best growth of the
organisms under investigation. The medium has to provide all necessary nutrients for
energy supply (catabolism), components for biomass formation (anabolism), and re-
ductants in connection with energy generation. The nutrients have to be present in
form and concentration that allows for rapid uptake and utilization by the cells. Fur-
thermore, support for anabolism has to be provided by providing amino acids, cofac-
tors or vitamins, as far as they are essential or not, present in the cell in amounts high
enough to obtain sufficient growth or production rates. The constituents are specific
for different biological groups and may be even strain specific. Selective and differen-
tial media are designed to support one organism and suppress others. One application
is the cultivation of genetically engineered strains with antibiotic resistance.
The medium determines the environmental conditions with respect to acidity
and buffer effectiveness, osmotic pressure, viscosity, and salinity. Some of the related
ingredients are not used by the cell but have to be present. In addition, practical
problems like foaming or sticking have to be checked. In addition, special ingredi-
ents can be necessary as chelating agents against precipitation of salts, protective
agents against toxic or inhibiting compounds, or inducers for product formation. The
choice of a medium is an important item in cost calculation especially for low value
products. Technical media therefore often have a different composition from lab me-
dia. Technical media derived from residuals, e.g., from food processing, have to be

https://doi.org/10.1515/9783110315394-003
42 | 3 Media – supplying microorganisms with an environment and building blocks for growth

checked against artificial media. Availability and handling issues are further points
for consideration.

3.2 Media design – starting from scratch

Before looking up recipes from handbooks, we try to translate basic knowledge about
the need of microorganisms into quantitative terms. In chemical engineering reaction
processes are quantitatively described by material balance equations applied to each
chemical compound appearing in the process. This requires knowledge of the stoi-
chiometry of each particular reaction. However, this molecular species balance is too
complicated for the moment, as we cannot follow all reactions inside the cell. Since the
number of atoms of each element is a conserved variable, the atomic species balance,
often called elemental balance, can be applied independently from specific reaction
schemes. In the following paragraph this idea is outlined for aerobic cultivation of
heterotrophic microorganisms as the most common case.
We start by drawing a system boundary around the biomass, in Figure 3.1. rep-
resented by a single cell. The molar amount of substances ‘Comp’ nComp [mol] going
into or out of the biomass per a given time interval ∆t has to now be considered. Molar
amounts as such are not a conserved quantity but allow us to set up the elemental bal-
ance. Therefore, all atoms forming part of the nutrients or products have to be added
up according to their relative quantity in the compound ‘Comp’. The stoichiometric
coefficients sE,Comp are defined as the molar ratio of Element ‘E’ in the compound
‘Comp’. The interval ∆t is the duration of a fermentation or the sampling time. Com-
pounds entering are the nutrients mainly present at the beginning of the cultivation,
while products and biomass itself are present at the end.
Writing down the balance equation for carbon (C) as an example yields:

nC = sC,S × nS + sC,O2 − sC,CO2 × nCO2 − sC,P × nP − sC,X × n X (3.1)

Substrate H2O
O2 CO2
Proteins
NH4+ Nucleic acids
Carbohydrates Product
Lipids
PO43- New
Biomass
SO42-
H2O

Fig. 3.1: The cell as catalyst, given compounds in the medium are converted to products and new
biomass. The exact metabolic pathways are not considered in this view.
 3.2 Media design – starting from scratch | 43

Note that the substances going out of the system are counted negatively. As we do not
allow any element to accumulate inside the system (new cells are counted as being
outside the boundary) the balance has to be zero:

nC = 0 (3.2)

The same holds for the nitrogen (N), phosphorous (P) and sulfur (S) balance:

nN = sN,NH4 ⋅ nNH4 − sS,P ⋅ nP − sN,X ⋅ n X = 0 (3.3)

nP = sP,PO4 ⋅ nP,PO4 − sP,X ⋅ n X = 0 (3.4)

This holds, assuming the product does not contain phosphorous:

nS = sS,SO4 ⋅ nNH4 − sS,P ⋅ nP − sS,X ⋅ n X = 0 (3.5)

In cases with different nitrogen sources like amino acids or products like proteins ad-
ditional terms come into the N and S balance equations. Nucleic acid as product leads
to an additional term in the P balance. To have a more generalized and compact no-
tation it is convenient to sum up the fractions of the element E over all nComp (= 10)
compounds:
nComp
∑ sComp,i ⋅ n i = 0 (3.6)
i=1

Having in mind that media can be designed with computer programs, an even more
compact vector notation here for the carbon balance yields:

sC ⋅ n = 0 (3.7)

The vector sC contains the values of the stoichiometric coefficients of the example:

s C = [6, 0, 0, 0, 0, 0, 0, 1, sC,P , sC,X ] (3.8)

The vector n contains amounts of substances:

n = [nS , nO2 , nNH4 , nPO4 , nSO4 , nH2O,in , nH2O,out , nCO2 , nP , n X ] (3.9)

The stoichiometric matrix s for all elements nE = 6 is composed as:

s = [sC , sN , s O , s N , s P , s S , s H ]T (3.10)

Finally, we can write down all stoichiometric information in the matrix equation:

s⋅n=0 (3.11)

For now we do not know a priori the carbon content of the biomass sC,X , nor the molec-
ular mass of biomass, which is obviously a virtual number.
44 | 3 Media – supplying microorganisms with an environment and building blocks for growth

For practical engineering masses are more convenient. These can be obtained by
substitution of the stoichiometric coefficients of the compounds by mass fractions
eE,Comp , here for element E in compound Comp:

sE,Comp ⋅ ME
eE,Comp = (3.12)
MComp

By substitution of the molar amounts of substance by mass and normalizing to volume

we get concentrations c.
The mass balance, e.g., for carbon, then reads:

eC,S ⋅ cS − eC,CO2 ⋅ cCO2 − eC,P ⋅ cP − eC,X ⋅ m X = 0 (3.13)

nq
∑ eC,i ⋅ c i (3.14)
i=1

For a better overview this sum can be split into terms representing the concentrations
at the start of fermentation and the concentrations at the end.
nComp,in nComp,out
∑ eC,i ⋅ c i − ∑ eC,i ⋅ c i = 0 (3.15)
in out

Employing elemental balances is already a good basis for medium design, but there
is still missing information. nE = 6 equations for the elements have been set up for
nF = 10 unknown concentrations. Some degrees of freedom in the system are left,
as we do not know how much carbon is actually directed into biomass and how much
into carbon dioxide and product. The same holds for nitrogen and its partitioning into
product or biomass.
Now a point is reached where additional stoichiometric data need to be included.
From batch experiments we know that the yield Y X,S is often close to 0.5 g/g for aero-
bic growth on glucose (without product formation). The background lies in ATP gen-
eration by respiration and the subsequent use of ATP for growth. This will be further
outlined in Chapter 4. We can now use this observation to get a preliminary empirical
value. Knowing this yield, we can fill in the first value in Table 3.1 to design a generic
medium. All information given in the balance equations represents relative relation-
ships. The most important decision is now to decide on a targeted amount of biomass.
For c X = 50 g/L new biomass as an example, cS = 100 g/L substrate is needed.
Up until now we have looked at the cell as being a molecular species. That is
of course not true. Furthermore, we do not know a priori the carbon content of the
biomass eC,X , nor the molecular mass of biomass, which is obviously a virtual num-
ber. Nevertheless, the elemental composition of the biomass can be measured and nor-
malized to one carbon atom giving a virtual molecular formula and molecular mass
as basis for engineering considerations. An example for an average microorganism is
given in Table 3.1 It contains elements being covalently bonded in macromolecules.
 3.2 Media design – starting from scratch | 45

Tab. 3.1: Elemental composition and molecular formula for an average generic bacteria ‘Virtuella
generica’; values are similar to E. coli.

Species Mass fractions eE,X [g/g] Empirical formula Molar

mass
C O N H P S [g/mol]
Virtuella 0.47 0.31 0.11 0.07 0.025 0.009 CH1.8 O0.5 N0.2 P0.02 S0.007 25.47
generica

Tab. 3.2: Generic medium to produce 50 g/L biomass without product formation.

Compound/ Purpose Rule Side Concentra- Amount n Comment

ion effects tion c [g/L] [mol/L]
C6 H12 O6 C-source, Yield, ATP Osmotic 100 0.556 yield C-source
energy balance pressure dependent
source
NH+4 N-source Elemental pH 7.07 0.392 NO−3 alternative
balance
PO3−
4 P-source Elemental Precipita- 3.72 0.0392 often limited in
balance tion nature
SO2−
4 S-source Elemental 1.32 0.0137 –
balance
K+ K-source Meas. 0.5 0.013 similary Na+ in
mass some MO
fraction
Cl− Counter Charge 8.68 0.245 alternatives?
ion balance

With these data we can further complete our medium composition under the speci-
fication to reach the biomass concentration of c X = 50 g/L as listed in Table 3.2 The
determination of a final biomass concentration is necessary as all equations derived
so far are relative to biomass formation, neglecting for now product formation. From
the N balance the ammonia concentration can be calculated:
eN,x ⋅ c x g
eN,NH4 ⋅ cNH4 − eN,x ⋅ c x = 0 → cNH3 = = 6.68 (3.16)
eN,H3 L
The phosphorous and sulfur balance reveals accordingly:
eP,x ⋅ c x g
eP,PO4 ⋅ cPO4 − eP,x ⋅ c x = 0 → cPO4 = = 3.75 (3.17)
eP,PO4 L
eS,X ⋅ c X g
eS,SO4 ⋅ cSO4 − eS,X ⋅ c X = 0 → cSO4 = = 1.35 (3.18)
eS,SO4 ⋅ c X L
K+ , as the main cellular inorganic cation, does not appear in the molecular formula
but is present in the cells with a fraction of about 0.01; according to mass balance it
has to be in the medium with 0.5 g/L.
46 | 3 Media – supplying microorganisms with an environment and building blocks for growth

Fig. 3.2: Graphical representation of Liebig’s law, where the staves are
the relative amount of the respective element. The maximum filling
volume is given by the smallest one, which, in an unbalanced medium
may not be the highest value.

Elemental balances have been adapted from plant fertilization. Carl Sprenger and the
chemist Justus von Liebig observed in 1828 that plants needed a balanced quotation of
N and P. In soil with high nitrate content only phosphate could bring better yields and
vice versa. This finding was called Liebig’s law of the minimum. It states that growth is
controlled by the scarcest resource, today called the limiting factor. Later this rule was
popularized by a graphical model (Figure 3.2., showing a barrel with staves of unequal
length, where the shortest one determines the holding capacity.
The next step is caring for the charge balance. Of course, the net charge has to be
zero. This approach allows us to find a combination of dry salts to be weighed in. Any
excess of charges in the medium is physically impossible. When using already defined
recipes, the charge balance is automatically zero as we weigh in neutral substances.
The charge balance is set up as:
nComp
∑ zC,i ⋅ n i = 0 , (3.19)
i=1

where z is the charge number of the respective compounds, e.g., zPO4 = −3. In our
self-designed medium the total charge is highly positive, so we have to find negatively
charged counter ions. A good candidate is chloride as Cl− . It is abundant in nature
but not directly a building block for cell growth and present in the cells only in small
amounts. In the intracellular space proteins and polysaccharides also act as negatively
charged counter ions. The charge balance also has to stay zero during the cultivation.
In our case it is not guaranteed that the cell takes positively charged ions up with the
same velocity as negatively charged ions. The most important example is NH+4 . Cells
take up the uncharged NH3 , leaving a proton H+ in the medium. Ammonia uptake
leads consequently to an acidification during the cultivation. This has to be compen-
sated by titration e.g., with NaOH, again with a drawback, namely increasing salinity.
All considerations so far are based on constant biomass composition and constant
stoichiometry of substrate uptake and product formation. In addition, practically un-
hindered uptake of the nutrients is assumed. This concept is called balanced growth.
 3.2 Media design – starting from scratch | 47

Tab. 3.3: Necessary mineral elements, their biological function, and content in E. coli.

Ion species Involvement in reactions and E. coli content Comments

structures eI,X [mg/g]
K+ Cofactor for certain enzymes (e.g., 0.12 Principle inorganic
pyruvate kinases) cation
Mg2+ Cofactor for certain enzymes (e.g., 0.08 Cell membrane
kinases); contained in ribosomes and
phosphate esters
Ca2+ Cofactor for certain enzymes especially 12.74 Cell membrane
exoenzymes (proteases)
Fe3+ Component of certain enzymes, 0.02 Reduction to Fe2+
cytochromes, DNA synthesis possible
Na+ Involved in membrane transport 0.07
Cl− Inorganic anion Essential
Trace elements [μg/g] < 100μg/g
Co2+ Part of vitamin B12 0.019 Synthesis B12 solely
bacteria
Cu2+ Cofactor for metalloenzymes (e.g., 0.038
cytochrome oxidase)
Mn2+ Cofactor for enzymes (e.g., PEP 0.018 Essential for all microor-
carboxylase, recitrate synthase) ganisms
Zn2+ Cofactor for metalloenzymes, 0.059 Essential for all microor-
especially RNA and DNA polymerases ganisms
MoO2−
4 Present in format dehydrogenase 0.006

Nevertheless, further optimization is necessary with respect to yield or growth veloc-

ity, a long and empirical process.
The elements mentioned so far are covalently bonded in the major cellular com-
ponents. Together with the cations K+ , Mg2+ , Ca2+ , and Fe3+ they are called macronu-
trients as they are required in considerable amounts. To get an idea of the quantities
to be fed, a look at the biological function is helpful. The quantitative contribution to
cell mass is given by a real analysis of Escherichia coli as listed in Table 3.3
Micronutrients, also called trace elements, are related to specific metabolic func-
tions like cofactors for specific enzymes or to maintain protein or cell membrane struc-
ture. The dosage is highly species dependent. The given sample analysis may not nec-
essarily reflect the real need of the cell but also the availability in the medium during
the experiment. Many companies have their own recipes with which they obtain good
results for a given purpose. Sophisticated use of micronutrients is also a means to
control metabolic fluxes to increase product formation. Examples are given in other
chapters.
48 | 3 Media – supplying microorganisms with an environment and building blocks for growth

3.3 Practical Application – a discussion of recipes from references

After setting up a theoretical minimal medium, a look at a commercial minimal me-

dia is advisable. Such media are used in labs to study microbial physiology. A popular
medium for use in labs is the M9 medium as given in Tables 3.4a, 3.4b and 3.4c in the
order of work flows. The recipe is given in the form of six different stock solutions to
be mixed for the ready-to-use medium. To prepare different stocks beforehand makes
handling easier, but is primarily a means to prevent mutual interactions between the
medium constituents during storage and sterilization, such as precipitation or chem-
ical reactions. Another point to take care about is that glucose solutions have higher
density than water (cS = 100 g/L →ρ = 1.04 kg/L; cS = 200 g/L → ρ = 1.08 kg/L).
The recipe is formulated on a mass per volume basis in the sense that the chemicals
of each stock solution are not to be solved in e.g., 1 L of water but have to be solved in
a smaller amount of water and then filled up to yield 1 L of solution. The same holds
for the other stock solutions. This approach is a tribute to easy pipetting and leads
to the precalculated glucose and salt concentrations in the final medium. An addi-
tional point is that 1 L of medium contains less than 1 kg of water. During cultivation
the density of the medium changes due to glucose uptake by the cells, what has to be
considered in feeding and sampling for precise balancing on a mass per mass basis.
The amounts of stock solutions are recommendations and do not influence the final
medium concentration.
Analysis of the medium shows that in the column ‘fraction from theoretical’ the
requirements for 2 g/L biomass with respect to yields and elemental balances are
fulfilled. Ammonia and sulfate are present in a slightly higher concentration than
required but to a reasonable extent. Nitrogen will be the limiting element for final
biomass concentration. Potassium, magnesium, and iron are present in unnecessar-
ily high concentrations. Phosphate is highly overdosed as it serves two purposes,
which are P-source and buffer. The sodium concentration is higher than required for
biomass formation as well due to its function as carrier of salinity.
For calculation on this basis a medium for 50 g/L biomass all salts necessary for
biomass formation have to be increased by a factor of 25. On the other hand, salts
that are not part of the biomass but are necessary for environmental conditions, e.g.,
sodium and chlorine for salinity, have to stay within given limits. Furthermore, our
goal is to avoid any waste, especially of phosphate; therefore it is necessary to increase
the N content of the medium to get a molar N/P ratio of 10 according to the empirical
formula of biomass composition. There are two degrees of freedom to go to higher
final concentrations, firstly to add larger volumes of the respective stock solutions to
prepare the final medium and secondly to increase the concentration of the respective
salts in the stock solutions. The first option is limited by the fixed final volume while
the second option is limited by the solubility of the salts in the stock solution and in
the medium.
 3.3 Practical Application – a discussion of recipes from references | 49

Tab. 3.4a: Composition of M9 medium (Cold Spring Harbor) modified for c Gluc,4 = 4 g/L glucose in
the final medium.

Stock ‘glucose’ mGluc,4 [g] M [g/mol] n Gluc,4 [mol] Remarks

100 mL

Glucose 20.0 180.2 1.109

Solve in 90 ml warm water bath, fill up to 100 mL
Use autoclave for sterilization
Stock ‘salts’ 1000 mL mSalt,4 [g] M [g/mol] nSalt,4 [mol] Remarks

‘5x’ Multiple of final concentration

Na2 HPO4 ⋅ 2H2 O 42.5 178.0 0.239 P-source, basic salt, buffer
KH2 PO4 15.0 136.1 0.110 Acid salt, used as buffer
NaCl 2.5 58.4 0.043 Adjusts salinity
NH4 Cl 5.0 53.5 0.093 N-source, acid salt
Dissolve in 800 mL, adjust to pH 7 using 4M NaOH, fill up to 1 L
Filter sterilize and store in the dark; autoclaving leads to ammonia evaporation
Stock ‘magnesium’ mMg,4 [g] M [g/mol] nMg,4 [mol] Remarks
100 mL
‘1 M’ Round molarity
MgSO4 ⋅ 7H2 O 24.65 246.5 1.00
Dissolve in 80 mL and fill up to 100 mL
Autoclave and store at room temperature
Stock ‘calcium’ mCa,4 [g] M [g/mol] nCa,4 [mol] Remarks
100 mL
‘100 mM’ Round molarity
CaCl2 ⋅ 2H2 O 1.47 147.0 0.01
Dissolve in 80 mL and fill up to 100 mL
Autoclave and store at room temperature

Stock ‘iron’ 100 mL mFe,4 [g] M [g/mol] nFe,4 [mol] Remarks

‘50 mM/100 mM’ Round molarity
FeCl3 ⋅ 6H2 O 1.35 270.3 0.05 Unstable
Citric Acid ⋅1H2 O 2.11 210.1 0.10 For complexation
Dissolve in 80 mL and fill up to 100 mL
Filter sterilize and store in the dark at 4 °C
Stock ‘trace elements’ mStock,4 [g] M [g/mol] nStock,4 [mol] Remarks
1000 mL
‘100x’ Multiple of final concentration
MnCl2 ⋅ 4H2 O 100 197.91 0.505
ZnCl2 170 136.29 1.247
CuCl2 ⋅ 2H2 O 43 170.48 0.252
CoCl2 ⋅ 6H2 O 60 237.93 0.252
Na2 MoO4 ⋅ 2H2 O 60 241.95 0.248
Dissolve in 800 mL, adjust to pH7, fill up to 1000 mL
Filter sterilize; more than needed to reduce errors from weighing accuracy
50 | 3 Media – supplying microorganisms with an environment and building blocks for growth

Tab. 3.4b: Dosing scheme for 1 L M9 medium.

‘Glucose’ ‘Salts’ ‘Magnesium’ ‘Calcium’ ‘Iron’ ‘Trace elements’ H2 O

[mL] [mL] [mL] [mL] [mL] [mL]
20.0 200.0 1.0 1.0 1.0 10.0 Add to 1 L

Tab. 3.4c: Final concentration in the medium.

Organics n Medium,4 c Comp,Medium g Compound / Fraction Remarks

or ion [mmol/L] [g/L] g biomass from the-
oretical

Glucose 22.2 4.00 2.0 1.00 Intentional

NH+4 18.7 0.34 0.17 1.18 Reasonable
−
HPO2−4 /H2 PO4 69.8 6.7 3.35 44.29 Buffer function
SO2−
4 1.0 0.1 0.05 1.78 Reasonable, reduction
Na+ 104.1 2.39 1.20 17.1 Salinity, nonhalophilic
K+ 22.0 0.86 0.43 4.3
Cl− 27.4 0.98 0.49 1.0 Salinity, counter ion
Mg2+ 1.0 0.02 0.01 5.21
Fe3+ 0.050 0.003 0.0014 14.0
Citric acid 0.100 0.019 0.0096 –
Ca2+ 0.100 0.004 0.002

Trace ele- nMedium,4

ments [µmol/L]

Mn2+ 5.1 0.28 0.14

Zn2+ 12.5 0.81 0.41
Cu2+ 2.50 0.16 0.08
Co2+ 2.50 0.15 0.08
MoO2−4 2.5 0.04 0.2

To get an overview of a possible route for arriving at an upgraded medium we can

make a thought experiment. Glucose is soluble in water up to 500 g/L. Choosing this
value for the stock solution yields for a factor of 2.5. The still missing factor of 10 can be
reached by increasing the added volume in the dosing scheme up to 200 mL. Dispens-
ing the 25-fold volume of the other stock solutions except ‘salts’ is possible as well, but
fills up the 1 L final medium already up to 52.5%. A slight reduction of ‘magnesium’ is
possible, while the amount of ‘iron’ could be strongly reduced. Phosphate has to be
increased only by a factor of 2.5 with respect to stoichiometry. That can be handled by
increasing the volume of stock solution ‘salts’ by the according factor to give a volume
to be added of 450 mL. Salinity still stays in the nonhalophilic range <1 mol/L. How-
ever, this answer to the problem brings us close to the total volume restriction. The
concentration of ammonium chloride has to be increased by an additional factor of 10
in the stock solution. However, this implies the risk of evaporation of ammonia into
 3.3 Practical Application – a discussion of recipes from references | 51

Tab. 3.5: A defined medium from literature for production of recombinant proteins by Bacillus.

Defined medium M final concentration

(Zhang et al. 1996) [g/mol] [g/L] [mol/L] Remarks
Glucose 180.16 112.50 6.24
(NH4 )2 SO4 132.14 16.50 0.12
K2 HPO4 174.18 1.50 0.009
Monosodium glutamate 169.13 7.50 0.044 Precursor for proteins
CaCl2 110.98 1.00 0.009
final concentration
[mg/L] [mmol/L]
MnSO4 ⋅ H2 O 169.02 2.00 0.0118
CuSO4 ⋅ 5H2 O 249.69 1.00 0.0040
ZnSO4 ⋅ 7H2 O 287.53 10.00 0.0348
CoCl2 ⋅ 6H2 O 237.93 0.80 0.0034
MgSO4 ⋅ 7H2 O 246.48 0.578 0.235
FeSO4 ⋅ 7H2 O 278.01 10.00 0.0360

the gas phase of the bottle. To circumvent this problem another N source like nitrate
or urea is an option, or we could feed ammonia separately during cultivation to avoid
too high concentration in the bioreactor.
To conclude this part of medium design, a look at a reference medium for techni-
cal applications is advisable, guaranteeing high concentrations. A successfully tested
medium is given in Table 3.5.
Solubility of a solute is defined as the analytical composition of a saturated solu-
tion, expressed in terms of the proportion of a designated solute in a designated sol-
vent, in our case salts to be dissolved in water. Values for the solubility of substances
are listed in chemical handbooks or on dedicated websites and are usually given for a
defined temperature and pressure in g salt per 100 g water or in g/L, e.g., the solubility
of NaCl is 360 g/L at 30 °C.
However, these values are valid only for pure salts. As we have mixtures of several
salts, the different cations and anions can interact mutually. Calculation is done by
applying the mass action law to the equilibrium between the solid and the dissolved
phase:
H2 O
CatsCatAnsAn (solid) 󴀕󴀬 sCat ⋅ CatsAn+ (liquid) + sAn ⋅ AnsCat− (liquid) (3.20)

The exponents are the charges of the respective ions. Take note of the charge balance
sCat ⋅ sAn+ − sAn ⋅ sCat− = 0. In further equations the charges are omitted for simplicity.
Introducing the equilibrium constant Kdiss,CatAn leads to:
sAn− sCat+
CatsCatAnsAn (solid) ⋅ Kdiss,CatAn = nCat ⋅ nAn (liquid) (3.21)
52 | 3 Media – supplying microorganisms with an environment and building blocks for growth

The dissolved ions on the right side of the equation also include ions with origins from
other salts in the medium sharing the same ions. In sum they shift the equilibrium to
the left side of the equation, thus increasing the risk of precipitation. This is called the
common iron effect. Checking the left side of the equation with a plausibility check
tells us that the position of the equilibrium does not directly depend on the amount
of solid salts in the medium. The term CatsCat AnsAn rather means a bit artificially the
number of moles in a liter of the undissolved salt. So it is a material constant and can
be combined with the dissociation constant Kdiss,CatAn to a new constant KSP called
the solubility product constant:

KSP,CatAn = nCat
sAn−
⋅ nAn
sCat+
(liquid) (3.22)

In principle, all possible combinations of ions after mixing the different salts have to
be checked against their KSP values. However, KSP takes no account of pH, ion activity,
and ionic strength and should be employed with care. In addition to the ions provided
by the medium, carbonate and hydrogen carbonate have to be considered as well, as
these compounds are formed from CO2 being produced by respiration. Depending on
the pH value of the medium the concentration of hydroxide ions is an issue. The for-
mation of sparingly soluble iron carbonate and hydroxide as well as the precipitation
of calcium carbonate in alkaline solutions have to be mentioned.
Depending on the pH and the ionic composition of the solution, phosphates of
iron, calcium and other heavy metals with low solubility can precipitate. Critical is,
for example, the formation of struvite, a sparingly soluble magnesium ammonium
phosphate (NH4 MgPO4 ⋅ 6 H2 O). Controlled by pH, temperature, and the presence
of other ions in solution such as calcium, struvite precipitates when the concentra-
tions of magnesium, ammonium and phosphate ions exceed its solubility product.
To prevent precipitation chelating agents can be employed. They form soluble com-
plexes with metal ions. Often used chelators are EDTA (ethylenediaminetetraacetate),
citrate, tartrate acid and gluconate. In our example citrate is used to complex iron ions.
The principle of complexation is also found as physiological performance in many mi-
croorganisms. Siderophores for example are excreted for acquisition of iron, which is
limiting in many environments due to its low solubility. Citrate – as used in our artifi-
cial medium – can also act as siderophore and is extracted by e.g., Aspergillus.
The selection of buffers is an important issue. For biological buffers significant
factors, among others, are: good solubility, pH range, buffer concentration, sensitivity
of pKa value to temperature and ionic strength, permeability through biological mem-
branes, no interaction with other components, nontoxicity, and low costs. Besides the
phosphate buffer system as in the medium described above, Tris (tris(hydroxymethyl)-
aminomethane), HEPES(N-(2-hydroxyethyl)-piperazine-N’-ethanesulfonic acid) and
MOPS (3-(N-morpholino)-propanesulfonic acid) are examples of ‘biological buffers’.
This term has become popular and means buffers for biotechnological use.
Defined media, where all compounds are exactly known and calculated before-
hand, are not always the best choice. For cost reasons and for supporting the cells
 3.4 Complex and technical media | 53

with necessary vitamins, cofactors or additional carbon and nitrogen sources, com-
plex media are often employed in industrial fermentations.

3.4 Complex and technical media

The first step is to supply the culture with growth factors. Besides vitamins (Table 3.6),
other compounds like specific amino acids, lipids, or nucleic acids may be essential as
they cannot be built up by the cells themselves. They bridge limiting metabolic steps
or help as protective agents; a first glance is given in Figure 3.3.
Complex media are used to supply microorganisms with necessary vitamins and
auxiliary amino acids. The goal is to allow most microorganisms to grow without
specifically selected medium ingredients. Lysogeny broth (LB), a nutritionally rich
medium, is primarily used for the growth of bacteria in the lab. LB media formu-
lations are used in molecular microbiology and have been an industry standard,
especially for the cultivation of Escherichia, for decades. There are several common
formulations of LB. Although they are different, they generally share a somewhat
similar composition of ingredients as given in Table 3.7

Tab. 3.6: Some of the frequently required vitamins.

Vitamin Amount [µg/g], Purpose Remarks

estimated

Biotin <1 Coenzyme for carboxylate reactions

(CO2 -fixation, TCC)
Vitamin B12 0.02 Rearrangement reactions (e.g., glutamate Can only be
mutase) produced by
some soil
bacteria
Folic acid <1 Coenzyme for carboxylate reactions
Nicotinic <40 Precursor for NAD+ and NADP+
acid (B3)
Pantothenic <20 Precursor for coenzyme A
acid (B5)
Vitamin B6 0.5 Precursor for pyridoxal phosphate and important
coenzymes in transaminase and amino acid
decarboxylation
Riboflavin (B2) 50 Precursor for FAD (coenzyme of flavoproteins)
Thiamin (B1) 2 Precursor of thiamine pyrophosphate (coenzyme
of decarboxylases and transketolases)
Vitamin K 1000 Precursor of menaquinone (electron carrier in
respiratory chain)
54 | 3 Media – supplying microorganisms with an environment and building blocks for growth

O2

Glucose TCC CO2

H2O

Product
Biomass
Precursor

Fig. 3.3: The strongly simplified structure of a microbial metabolism for medium design; possible
additives in the medium are for overcoming bottlenecks, marked with the brownish double ellipse.

Tab. 3.7: Composition of a LB medium.

Compound Amount [g/L] Purpose Remarks

Yeast extract 5 Diverse organic compounds, vita- Autohydrolysate of yeast cells

mins (B), trace elements
Tryptone 10 Peptides, peptones, source of Casein, hydrolyzed by pro-
amino acids teases trypsin, pepsin
NaCl 5 Transport, osmotic balance Varies depending on required
salinity
Glucose 1 Easy to use energy and carbon Or another C source
source

This brute force method is easy to use in labs and brings most organisms to growth, but
it is not well enough defined for quantitative physiological studies as uptake of spe-
cific substances cannot be followed by analytics. LB has its serious drawbacks for large
scale production. Substances such as yeast extract automatically imply significant dif-
ferences in composition between batches. It is obvious that the production of cell mass
on other cell mass (yeast extract) and processed proteins will not be economically fea-
sible. Other problems include lack of ability to be upgraded for high biomass concen-
trations and technical problems like foaming. Nevertheless, in commercial production
yeast extract and tryptone are also occasionally used in smaller amounts for supply of
essential amino acids and vitamins. In some bioplants yeast extract is stockpiled for
a year to be ensure against changes between consecutive production runs.
One of the most widespread technical media is sugar cane molasses (Figure 3.4).
It is the end product of the sugar manufacturing process: once no more sugar can be
crystallized from the raw crop, the residual product is molasses, still containing up to
20% of total sugar in the cane. The dark colored highly viscous liquid contains up to
500 g/kg sugars so it is ideally suitable as fermentation medium. Molasses, also called
 3.4 Complex and technical media | 55

Fig. 3.4: Molasses is a viscous and brownish liquid;

pigments are from heat treatment during extraction of
sugars.

‘blackstrap’, contains a rich variety of trace elements and vitamins. The composition
varies depending upon region of origin. An analysis of a commercially available mo-
lasses is shown in Table 3.8 The density is higher than the density of water, so values
are given on the basis of mass rather than volume.
The carbon source is mainly sucrose. This is clear as molasses is a plant derived
product. Due to hydrolytic activities (sulfuric acid) during sugar production sucrose
is partly hydrolyzed to glucose and fructose, a mixture called invert sugar. This term
evokes the change of the direction of optical rotation during this process. Not all mi-
croorganisms can take up sucrose because of missing invertase activity to split su-
crose. Yeast develops high activities during cultivation on molasses. Invertase for food
application is actually produced by yeasts.
The high sugar concentration in the molasses makes it necessary to care for the
osmotic pressure in the medium. Osmotolerant yeasts can grow in media with concen-
trations of 40–70% sugar. Therefore, it is possible that contaminants can come into the
process from wild yeasts. On the other hand, the osmotic pressure of the solution is
so high that our production strains suffer from suboptimal conditions. For batch pro-
cesses molasses has to be diluted down to sugar concentrations of about 200 g/L. In
Chapter 8 other process policies will be shown that use concentrations even as high
as 500 g/L.
Even at the first glance, it can be seen that some minerals are not present in
amounts sufficient to support balanced growth up to high cell densities; see exer-
cises. Therefore, additional nitrogen source ammonia is usually employed, but amino
acids or proteins from other residuals can also be considered. This could be for exam-
ple cheese whey or corn steep liquor (Table 3.9). Some small factories produce ‘organic
yeast’ only based on sustainable raw materials, where nitrogen source hydrolysates
from proteins are an option. Trace elements like selenium are bonded to the organic
matrix by the yeast cell thus increasing their bioavailability for humans in deprived
areas. Cobalt is available for use by animals only after being bound e.g., to vitamin
56 | 3 Media – supplying microorganisms with an environment and building blocks for growth

Tab. 3.8: Composition of typical sugarcane molasses on a mass per mass basis.

Constituent Unit Value

Dry matter % (g/100 g) 80
Total sugars % 50
of which sucrose % 67
of which inverted sugar % 33
Nonsugar organic matter % 10
N containing nonsugar % 4
amino acids, peptides
measured as N ⋅ 6,25
of which can be taken up by yeast % 20
N free nonsugar % 6
hemicellulose, organic acids
Mineral elements %
Sodium 0.2
Potassium 3.5
Calcium 0.6
Magnesium 0.4
Chlorine 1.3
Sulfur 0.5
Phosphorous 0.1
Trace elements mg ⋅ kg−1
Selenium 0.02
Cobalt 0.5
Copper 9
Manganese 20
Zinc 10
Iron 200
Vitamins μg ⋅ g−1
Biotin 2.0
Pantothenic acid (vitamin B5) 20.0
Inositol 4000
Thiamine (vitamin B1) 1.8

B12 by yeasts or bacteria. Growth factors including vitamins are found in adequate
quantities in numerous natural media. Nevertheless, they help to get the best growth
rates and yields.
After harvesting the yeast biomass, the brown pigments have to be washed from
the cells. Together with other residuals in the used-up medium this creates a con-
siderable wastewater problem, showing that bioprocesses are not automatically envi-
ronmentally friendly in all aspects. For cultivation processes, which require a higher
quality, high-test molasses is employed. The process follows the same pretreatment
of sugar cane, but the steps of acidification, heating up and sugar crystallization are
 3.4 Complex and technical media | 57

Tab. 3.9: Composition of corn steep liquor (expressed on a 100% dry matter basis).

Constituent Unit %=[g/100 g] Value

Dry matter % 53
Proteins (N x 6,25) % 45
Amino acids and
peptides
Fat % 10
Ash % 18
Mineral elements %
– Sodium 0.33
– Potassium 2.25
– Calcium 0.11
– Magnesium 0.66
– Phosphorous 3.12
Carbohydrates % 37
– Lactic acid 20
– Phytic acid 10

skipped. This medium is a clear, light brown syrup with a much better controlled com-
position than blackstrap molasses. Sugar content is up to 80%. The gentle processing
makes addition of dry granular yeast necessary to care for invertase activity.
A complex medium with high nitrogen content and high nutritional value is corn
steep liquor, a residual from the first washing and extraction step in the process of
starch production from corn (Table 3.9). It is a brownish highly viscous fluid.
The high lactic acid content caused by fermentative activity during produc-
tion is typical. Phytic acid acts as phosphate storage in many plants. The formula
C6 H18 O24 P6 shows the high phosphorous content thanks to six phosphate residues
in the molecule. To guarantee availability, phytase activity has to be provided by the
strains to be cultivated. On the other hand, corn steep liquor is rich in amino acids
including all possibly essential ones making it a suitable medium or as adjunct to
molasses for organisms that are auxotrophic in some amino acids.
With a similar function in media design, cheese whey can be employed. The 6%
dry matter contains mainly lactose (4%), proteins (1%) and minerals (P, Ca, Na, K). Of
specific value are the high concentration of vitamins and cofactors including Vitamins
A, B12 , and C, as well as choline and riboflavin. The focus of usage in fermentation
lies in lactic acid bacteria and other anaerobic processes but also yeast biomass pro-
duction (Candida). Lactose has to be hydrolyzed preceding cultivation in some cases
where the organisms do not show lactase activity.
Looking for other potential complex technical media, glycerol is a candidate. It
is a highly abundant byproduct of biodiesel production. This was inspired by the use
of glycerol as a cheap carbon source for fermentation. Production of 1,3-propanediole
58 | 3 Media – supplying microorganisms with an environment and building blocks for growth

(anaerobic) is of industrial importance, and succinic acid is promising and under de-
velopment. Lignocellulose in the form of straw or wood chips is the most abundant
raw material for possible use as renewable resource. However, prior to fermentation
hydrolytic pretreatment is necessary. Otherwise, the lignin network protects the cellu-
lose against microbial attack. In fact, that is one of its biological functions. That is ob-
vious when observing that wood decay takes months under natural conditions. White
rot fungi degrade lignocellulose in nature, where in the first stage cellulose (white) is
left over for a while. This has evoked some scientific interest, as cellulose is a valuable
material for other purposes. The idea is to use a medium without specific ions being
cofactors for cellulases. This idea of influencing cell activity by targeted ion depletion
in the medium unfortunately did not work out in this case but is an applied option
for other processes. Enzymatic hydrolysis of straw with the lignin degrading enzyme
laccase derived from the mentioned rot fungi is under investigation at the demonstra-
tion scale. Therefore, the usage of straw and wood as a fermentation medium meant
to convert this cheap raw material into high value products or liquid fuels is left for
the future.

3.5 Additional aspects – mutual interactions between cells and

medium

Up to this point we considered mainly medium composition under the condition of a

constant elemental biomass composition and therefore the final biomass concentra-
tion. Nevertheless, the other way round is also true: media compounds also influence
cell composition. This can be the accumulation of intracellular storage compounds or,
especially for eukaryotes, changes in macromolecular cell composition, which has to
be considered in media design.
As persons responsible for process development we are interested not directly in
elemental but in macromolecular composition of the biomass. In fact, the elemental
composition is a consequence of the macromolecular composition. All organisms have
a specific content of proteins, nucleic acids, carbohydrates, or lipids, with which they
adapt to a given environment. That holds of course also for a higher level of detail,
meaning e.g., the content of specific enzymes. The other way round is the idea to shift
the macromolecular stoichiometry by a targeted deprivation or surplus of specific nu-
trients. A typical macromolecular composition is given in Table 3.10, with eMm/X being
the mass fraction of a molecular compound Mm in cell dry mass. Lipids as a structural
part of each cell are listed here as well. Of course, there are differences with respect to
the group of microorganisms and especially the growth rate.
Now we try to calculate back the macromolecular composition (Table 3.11) to the
elemental composition. As a first attempt, representative molecules are chosen for
 3.5 Additional aspects – mutual interactions between cells and medium | 59

Tab. 3.10: Typical macromolecular composition of microorganisms.

Species Mass fractions eMm,X [g/g] on dry mass basis

Protein Nucleic acids Carbohydrates (Phospho)- Storage Others (ions,
(RNA/DNA ≈ 7) Lipids compound metabolites)
Virtuella 0.55 0.20 0.15 0.1 0.00 0.03
generica

Tab. 3.11: Elemental composition of macromolecules based on representative mono- and polymers.

Empirical formula Molecular mass Based on

M [g/mol]
Proteins C1489 H2403 N409 O453 S10 33,603.42 Glucokinase Bacillus subtilis
Nucleic acids –C39 H49 O25 N15 P4 – 1,251.79 RNA chain of C,G,A,T/U
Carbohydrates –C6 H10 O5 – 162.14 Glucose -H2 O
(Phospho)-Lipids C42 H79 O10 P 775.05 Dioleoyl-phosphatidy-glycerol
Others

Tab. 3.12: Comparison of the previously assumed elemental composition and the calculated elemen-
tal composition from the macromolecular composition.

Species Mass fractions eE,X [g/g] and eE,Mm [g/g]

C O N H P S
Virtuella generica 0.47 0.31 0.11 0.07 0.025 0.009
Proteins 0.532 0.216 0.170 0.072 0.000 0.010
Nucleic acids dAGCT 0.374 0.320 0.168 0.039 0.099 0.000
Carbohydrates 0.444 0.493 0.000 0.062 0.000 0.000
(Phospho)-Lipids 0.651 0.206 0.000 0.103 0.040 0.000
Others
Calculated Virt. gener. 0.499 0.277 0.127 0.067 0.024 0.005

which the empirical formula is known. These monomers and polymers could be de-
fined further by data from public databases or references.
From this table the elemental composition can be estimated as shown in Table 3.12
The calculated empirical formula of the virtual microorganisms Virtuella gener-
ica turns out to be quite close to the ‘measured’ elemental formula of this species.
This is only a crude estimate but justifies our view that the elemental composition
is a projection of the macromolecular composition. For complex media such calcula-
tions can give a first basis for possible yields of biomass from a given medium. This
view can also be applied vice versa. Changing the elemental composition by changing
the medium will possibly change the macromolecular composition and therefore the
physiological state of the cells. This is especially true for species that can accumulate
storage material. One example is the accumulation of bioplastic PHB by some bacte-
60 | 3 Media – supplying microorganisms with an environment and building blocks for growth

ria during nitrogen starvation. The cells are then no longer in a position to synthetize
proteins and nucleic acids. Catabolic reactions can nevertheless proceed allowing the
cells to produce ATP and accumulate carbon rich compounds. Besides accumulation
of storage material, prokaryotes are quite constant in their composition. A fixed ra-
tio of lipids, proteins and polysaccharides is necessary to build up the cell membrane
and cell wall. The cell machinery will work properly only for an optimal protein to nu-
cleic acid ratio. Eukaryotes like yeast (see running case study) can adapt to a changing
supply of different nutrients with e.g., changing protein content.
The actual concentrations of many compounds can influence growth velocity with
respect to limitation or inhibition. In other cases, two or more compounds can posi-
tively interact in supporting growth as we have sketched for essential amino acids and
growth factors. These aspects are discussed in the next chapter on kinetics. Optimiza-
tion in media design is discussed under process strategies and in the case studies.
The overall approach to design media is summarized in Figure 3.5. It is important to
understand that thinking always starts from the top and not from the bottom with ap-
plication of unaudited recipes.

Basic Idea:
What does the
cell need?

Constraints:
Stac Need for
Metabolic
Environment macromelecules
pathway

Simplificaon:
Simplificaon:
pH, salinity, Lumped
Elemental
osmolality metabolic
balance
structure

Closed Stoichiometry
Rule of
elemental where possible
balances thumb: Yields
aerob, anaerob

Fig. 3.5: The standard approach to medium design; the general idea to give the cell what it needs is
broken down to different aspects of quantification.
 3.6 The Good, the bad and the ugly – microorganisms as products | 61

3.6 The Good, the bad and the ugly – microorganisms as products

Now we have nearly all intellectual means to start a bioprocess and can start thinking
about products. The best unique selling point are microorganisms as there is no other
technical process to deliver them for different purposes and different markets. Due
to their ability for autocatalysis they can be the commercial product itself. This may
happen on a small scale when a lab purchases interesting strains from a strain bank
or on a large scale for industrial application. Different application areas are listed in
Table 3.13.

Tab. 3.13: Applications for which living microorganisms and cell lines are traded as products.

Purpose of employment Organisms

Samples from strain banks Nearly all
Starter culture Yeasts, Lactobacillus
Protein source, food and feed Yeasts, bacteria, microalgae
Agrochemicals Bacillus thuringiensis
Plant propagation Plant cells, e.g., cyclamens
Cell based therapy Animal cells
Bioweapons Anthrax

Food and feed processing is a classic field where microorganisms are employed. The
bacteria, yeasts or fungi are customarily not propagated directly in the food factory but
are purchased from professional suppliers. The most popular example are bakeries,
where the yeast is bought from yeast factories. The same holds for vintners and brew-
eries. Similarly, Lactobacilli (and others) are necessary for preparation of dairy prod-
ucts like yogurt, cheese, sourdough, or fermented vegetables. The inoculum for start-
ing the food fermentation process is called the ‘starter culture’ (fermentation starter,
in some branches ‘mother’). Some of them, like fungi to prepare Kombucha tea, are
even available for private application. To guarantee a high level of reproducibility and
quality the food manufacturer relies on the experience of the deliverer of the starter
culture with respect to genetic stability and physiological activity. The microorgan-
isms are not only active as biocatalysts in the food but contribute also to taste, smell
or nutritional value.
Apart from a food matrix, microorganisms are used directly as food or food sup-
plement. Since the 1960s, single cell protein (SCP) from edible microorganisms was
assumed to contribute substantially to food and feed supply for mankind. Feed and
food yeasts (Candida, Pichia) were the focus. As substrate hydrocarbons or methanol
were foreseen, as fossil oil reserves seemed to be exhaustless. Increasing oil prices
and other economic and political constraints made this development invalid. The
trend of producing energy from biomass with the aid of microorganisms shows that
62 | 3 Media – supplying microorganisms with an environment and building blocks for growth

value added chains can even be turned to the opposite direction, depending on mar-
ket needs and available resources. Nevertheless, the best example for a viable SCP
process for animal feed (ICI Pruteen) led to the development of continuous culture
in large scale of 1500 m3 . There are ongoing activities to produce high value proteins
in the form of edible microorganisms from sugar, lignocellulosic materials or other
agricultural residues. The main example for human novel food is ‘Quorn’ (after Leices-
tershire village of Quorn) with mycoprotein as its main ingredient. It is produced by
the Fusarium venenatum fungus in a continuous fermentation process. While Quorn
was developed in the 1980s and has been available as retail product since 1993 the
current trend towards vegetarian diets supports its prevalence (market value of over
100 million €/$) as it is a perfect substitute for meat.
Treatment of food by microorganisms or producing microorganisms for food on
agricultural products is highly welcome as a refinement process. Unfortunately, such
processes are competitors for arable land in the bioenergy/food nexus. As food pro-
duction seems to be limited and fossil carbon sources are even soon to be exhausted,
production of microorganisms based on organic substrates to contribute to the expo-
nentially growing human protein demand is hopeless. A way out of this dilemma is to
consider sun energy, which is technically used to produce electrical energy and con-
secutively hydrogen. This route of energy production is meant to be available in ex-
cess in the future. On the chemical side, synthesis of fuels from hydrogen and carbon
dioxide is called ‘power-to-fuel’ technology. Consequently, on the biotechnological
side, growing microorganisms assimilating H2 and CO2 to produce CH4 following this
route is a recent challenge. This is an example that others besides oxygen and carbon
dioxide can serve as educt and product. In the next step, conversion of the gaseous
substrate methane to protein by microorganisms may be a viable route to help ensure
global food security (e.g., by company Calysta). Note that this route is exactly diamet-
rically opposed to the attempts to make electrical energy or hydrogen by microorgan-
isms from organic matter.
The examples given above are meant to serve as beneficial processes fulfilling
human needs, but humans have also learned to misuse microorganisms as weapons
aimed at each other. The ability of microorganisms to propagate and produce ex-
tremely strong toxins is the kinetic basis of many fatal diseases. This means that since
antiquity people have used microorganisms as weapons even without knowing the bi-
ological background. Even 3,000 years ago the Hittites employed infected cattle to sap
the food base of their enemies. Ancient people contaminated the drinking fountains of
their enemies with decaying corpses. In the Middle Ages infested bodies killed by the
plague were thrown over city walls. During the settlement of Europeans in America
the indigenous population was decimated by smallpox epidemics. In North America
some cases are reported where European solders are accused of infecting native tribes
deliberately with smallpox e.g., by deliberately giving them things like infested blan-
kets. Since the nineteenth century, when biotechnology was put on a rational basis,
 3.6 The Good, the bad and the ugly – microorganisms as products | 63

not only the fight against diseases but also the development of biological weapons
was the focus of research driven by political demands.
Bioweapons have the potential to kill people but to leave material values like
buildings unaffected. Botulinum toxins, known from pharmaceutical cosmetics (Chap-
ter 1), have been produced as bioweapons in several countries starting before the Sec-
ond World War. It is quite unstable under outdoor conditions, so that an affected area
can be entered already after two days. A recent study said that 1 g toxin distributed
by infested packaged milk could potentially kill 100,000 people, mainly children.
In contrast, in the twentyfirst century research begun in the area of ‘nondeadly’
weapons. The idea is to destroy specific materials, e.g., by microbes degrading fuel
reserves. Plant and animal toxins are also under inspection by militaries. The military
logic here is that such weapons are currently not covered by the Biological and Toxin
Weapons Convention (BTWC). On the other hand, they have some inherent limitations
making them, generally speaking, incalculable. This has prevented excessive so far.
Spreading of an epidemic cannot be strictly limited to a given area once broken out.
This causes a potential danger for the instigator’s own people. The development of
specific sensors detecting the toxic compounds under such scenarios is the next logic
step in the arms race. The onset of the impact will take some time, which enables
the opponents of war to take counter measures. Furthermore, it is regarded that in-
fected soldiers already facing death are specifically dangerous. The basic question
of how the microorganisms can distinguish between their own population and the
enemy’s is treated on the genetic level. Specifically designed bacteria, so called target-
delivery systems, are conceivable, which attack only people carrying specific genes.
The attacker’s own people could be protected by specific antibiotics. Such ethnic
weapons are discussed as the most appalling scenario. Horrible is the imagination of
bioweapons in the hand of terrorists (bioterrorists) not caring for collateral damage
or ethics at all.
The Center for Disease Control and Prevention (CDC) published a list (out of a po-
tential 200) of the ‘dirty dozen’ biological warfare agents (Table 3.14). These bacteria,
viruses, or toxins are especially dangerous because of their high lethality in conjunc-
tion with easy transmission and spread.
It is highly desirable that engineers working on bioprocesses are beware of these
potential dangers for mankind and actively take part in the political discussions to
ban such horrible developments.
64 | 3 Media – supplying microorganisms with an environment and building blocks for growth

Tab. 3.14: List of biological warfare agents.

Pathogen Disease Release Incubation

Bacillus Bacterium Inhalation anthrax Spores Aerosols
anthracis
Yersinia pestis Bacterium Pneumonic plague Vegetative cells Aerosols,
person to person
Francisella Bacterium Tularemia (Rabbit Vegetative cells Aerosols
tularensis Fever)
Brucella suis Bacterium Brucellosis Vegetative cells Aerosols
Coxiella burnetii Bacterium Q fever Vegetative cells Aerosols
Burkholderia Bacterium Glanders Vegetative cells Aerosols
mallei
Variolavirus Virus Smallpox Virus Aerosols, person
to person
VEE-Virus Virus Venezuelan equine Virus Aerosols
encephalitis
Marburg-Virus Virus Marburg fever Virus Aerosols, person
to person
Clostridium Bacterium Botulism Toxin (protein) (known Ingestion, food/
botulinum from cosmetics) water
Ricin Plant Ricin intoxication Toxin (protein) Food/water
Staphylococcus Bacterium SEB intoxication Toxin (known from Food/water
aureus food poisoning)
Trichothecene Fungus Trichothecene Toxin Food/water

3.7 Exercises, questions and suggestions

1. Compare the list of vitamins (Table 3.8) with the list in the introduction (Fig-
ure 1.3 (b)).
2. To produce ethanol with a newly developed strain of Clostridium aceto-
butylicum some experiments for process optimization in pilot scale (VR =
100 L) shall be designed. You already know that at high ethanol concentra-
tions of cP,end > 92 g/L cell lysis occurs. Design a balanced medium with
respect to glucose as substrate, NH4 CL as N source and KH2 PO4 as P source.
The elemental mass composition of the biomass was determined as

eE = [eC,X eO,X eN,X eH,X eP,X ] = [0.5 0.34 0.09 0.05 0.02] .
 3.7 Exercises, questions and suggestions | 65

1. Practically all vitamines for humans are also used in media, at least for some
strains.
2. The first step is to collect some additional assumptions. The process is anaerobic,
so we expect a biomass yield of y X,S = 0.05. Per 1 mol of glucose processed on the
fermentative pathway 2 mol of ethanol and 2 mol of CO2 will be produced. This
allows for calculation of the amount of necessary glucose to reach the maximum
ethanol titer:

cP,max = 92 ⋅ g ⋅ L−1 → nP,end

= 2 ⋅ mol ⋅ L−1 → nS,0,Eth (3.23)
−1 −1
= 1 ⋅ mol ⋅ L → cS,0,Eth = 180 ⋅ g ⋅ L

According the yield the final biomass concentrations c X,end = 10 g L−1 :

c X,end = c X,0 + y X,S ⋅ cS,0 = 1 ⋅ g ⋅ L−1 + 9 ⋅ g ⋅ L−1 = 10 ⋅ g ⋅ L−1 (3.24)

The 9 g ⋅ L−1 newly built biomass also needs a carbon source. We assume that this
is taken from pyruvate or glucose respectively without further net CO2 turnover:

6 ⋅ MC 72
eC,X ⋅ ∆c X = eC,Glu ⋅ cGlu,X = ⋅ cGlu,X = ⋅ cGlu,X
MGlu 180
→ cGlu,X = 11.25 ⋅ g ⋅ L−1 (3.25)

Finally, we need about 180 + 11 = 191 g L−1 of initial glucose concentration.

To provide enough N-and P-source the calculation is:

MN 14
eN,X ⋅ ∆c X = eN,NH4CL ⋅ cNH4CL = ⋅ cNH4CL = ⋅ cNH4CL
MNH4CL 53.5
→ cNH4CL = 3.10 ⋅ g ⋅ L−1 (3.26)

MP 31
eP,X ⋅ ∆c X = eP,KH2PO4 ⋅ cKH2PO4 = ⋅ cKH2PO4 = ⋅ cKH2PO4
MKH2PO4 136
→ cKH2PO4 = 0.79 ⋅ g ⋅ L−1 (3.27)
4 Kinetics – finding quantities for bioprocess
reactions
One of the unique selling points of microbial reactions is the formation of biomass and
complex high value products from medium compounds. These reactions take place al-
most entirely inside the cells. For design and assessment of bioprocesses the velocity
of the reactions and intrinsic stoichiometry has to be understood. A fundamental con-
cern of this chapter is to simplify complex but basically known biological relationships
and to transform them into manageable equations. Similar to the elemental balances
in medium design, strict constrains like intracellular balances are helpful and often
sufficient to set up macroscopically observable relations. Generally speaking, reac-
tion engineering principles are applied to the metabolic network. As consequence of
simplifying, a priori unknown parameters have to be defined and measured case by
case. The following chapter starts from formulation of enzymatic reactions as the ba-
sic units of cellular metabolism. The second step is to find kinetic equations for the
velocity of substrate uptake and growth. Finally, reasonable coefficients describing
stoichiometric relations between the different pathways will be formulated.

4.1 Kinetics – the scaffold of reaction engineering

In physics and engineering, kinetics (Greek: ‘kinesis’ = movement) is a term for de-
scription of the relationship between the motion of bodies and its causes, namely
forces. By analogy, reaction kinetics is the study of rates (= velocity) of chemical re-
actions. The forces here are material concentrations. Kinetics is the scaffold for un-
derstanding and designing systems in reaction engineering. Some macroscopically
observable quantities can be directly extracted from measured data.
Reaction rates R measure the number of reacting moles ∆mol or masses ∆m of a
substance Comp in a given time interval ∆t e.g., the unit time. Related to volumetric
concentrations, they are given here as QComp [mol ⋅ L−1 ⋅ h−1 ] or as RComp [g ⋅ L−1 ⋅ h−1 ].
Growth and product formation turnover rates are an important measure for the effi-
ciency or intensity of a bioprocess. The differences and the resulting rates are often
calculated such that they become positive values. So, we speak of ‘product forma-
tion rate’ and ‘substrate consumption rate’, making both positive even though glu-
cose concentration decreases. A second important measure is the amount of biomass
or product being formed by a given amount of substrate used. This is measured as the
biomass yield Y X,S = ∆m X /∆mS or product yield Y X,P = ∆mP /∆mS . Yield is a consid-
erable contributer to the cost structure of the production process. Here again positive
values are pursued despite decrease or increase of the involved concentrations. Vol-
umetric productivity PV [g ⋅ L−1 ⋅ h−1 ] measures the amount of product that has been
formed in a reactor with the working volume VR for the duration of the fermentation

https://doi.org/10.1515/9783110315394-004
 4.1 Kinetics – the scaffold of reaction engineering | 67

120
110 cX,0
100
90
80
cX [g⋅L-1], cS [g⋅L-1]

70 ΔcS
60 Δtc
cX,harvest
50
40
30
20 Inoculum cX,0 ΔcX
10 Δtc tc,harvest
0
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

Culvaon me [h]

Fig. 4.1: Measurements of biomass and substrate concentration of a typical batch process. The dif-
ferences are counted positively from the arrowhead backwards.

Tab. 4.1: Calculated values for the experiment shown in Figure 4.1.

R X [g ⋅ L−1 h−1 ] RS [g ⋅ L−1 h−1 ] Y X,S [g ⋅ g−1 ] PV,X [g ⋅ L−1 ⋅ h−1 ]

∆t c = (5–4) h 13.0 − 10.5 = 2.5 86.5 − 81.0 = 5.5 2.5
5.5 = 0.455 13.0 − 10.5 = 2.5
∆t c = (10–9) h 36.0 − 29.7 = 6.3 43.6 − 29.2 = 14.4 6.3
14.4 = 0.438 36.0 − 29.7 = 6.3
42.8
∆t c = (12.5–0) h 100.0 = 0.428 47.8 − 5.0 = 42.8

process as PV [g ⋅ L−1 ⋅ h−1 ] = ∆m/∆t/VR . For batch processes it is formally equal to the
production rate but interpretation is in done in a sense of efficiently using the volume
of the reactor.
A batch experiment based on the medium developed in the last chapter is shown
in Figure 4.1; characteristic numbers can be taken as a rough evaluation from the data
sheet in Table 4.1.
Biomass formation rate, or in fact growth rate R X , increases during the middle
phase of cultivation and is obviously low at the beginning and even negative at the
end. Substrate uptake rate RS increases as well in the same interval. The yield coef-
ficient Y X,S is quite constant showing only a small decrease. For the whole cultiva-
tion the overall yield from inoculum to harvest is Y X,S = 0.428 g ⋅ g−1 , lower than
the expected 0.5 g ⋅ g−1 . Volumetric productivity PV,X between the start of the fer-
mentation and the harvesting time at the point at highest biomass concentration is
42.8 g ⋅ L−1 h−1 . The major contribution happens in the late middle phase of the pro-
68 | 4 Kinetics – finding quantities for bioprocess reactions

cess. Beyond these practical characteristic numbers, we have the ambition to describe
the underlying reaction mechanisms and understand them as a function of process
conditions. Some aspects are given in the following paragraphs laying the foundation
for rational process design.

4.2 Enzymes as basic components – determining kinetics

Before starting with biological systems, we first refer to the mass action law, which is
the fundamental approach in chemical reaction engineering.
Firstly, we consider a simple reaction equation:

aA + bB 󴀕󴀬 cC + dD (4.1)

The equilibrium is found for a specific set of concentrations:

cCc ⋅ cDd
Kequ = (4.2)
cAa ⋅ cBb

Here Kequ is the equilibrium constant. The background is the idea that the reac-
tions are due to collisions of reactant species involved. The frequency with which the
molecules collide depends upon their concentrations, or more precisely on their ac-
tivity being dependent inter alia on temperature. The probability of a collision of two
different reactants is given by the product of their respective concentrations. Whether
a reaction takes place or not depends on the activation energy. This encourages us to
formulate volumetric reactions rates R [mol ⋅ L−1 h−1 ] as:

Rforth = k AB,CD ⋅ cAa ⋅ cBb (4.3)

Rback = k CD,AB ⋅ cCc ⋅ cDd (4.4)

The parameters k X,Y are the reaction constants e.g., from the compound X to Y. The
unit depends on the order of the reaction. In cases where a single compound disso-
ciates into two compounds, it has the dimension of a concentration. The equilibrium
is reached when forward and back reaction are equal. The mass action law will be
applied to enzymatic reactions in the next paragraph.
Enzymes catalyze almost all reactions in the cell. Isolated enzymes are also used
in bioprocesses in vitro as effective catalysts. Substrate molecules bind to the enzyme
at the so called active site, where the conversion occurs. Thanks to their spatial struc-
ture enzymes exhibit a much higher substrate specificity compared to inorganic cata-
lysts. The widely known ‘lock and key’ model represents the exact fit of the substrate
molecule onto the active site excluding other molecules with a different structure. Fi-
nally, the products are released and the enzyme is ready to bind to the next substrate
molecule. Biochemical reactions are proceeding far from their respective equilibrium.
 4.2 Enzymes as basic components – determining kinetics | 69

Substrate Enzyme-Substrate- Product

Complex
kE+S,ES kES,P

kES,E+S

Enzyme Enzyme

Fig. 4.2: Schematic representation of an enzymatic reaction process. The ‘pocket’ of the enzymes,
into which the substrate fits, is the binding site and shown here brighter than the enzyme itself.
The small hook represents the reaction site. The arrows indicate direction of the respective reaction
step.

So, a backward reaction from the product to the substrate is usually negligible. Fig-
ure 4.2 shows a representation of an enzymatic reaction process and will be used to
set up the reaction equations following the mass action law.
For many enzymatic reactions it is assumed that the first step, the binding of the
substrate S to the enzyme E, is reversible, while the reaction step and release of the
product is irreversible leading to the reaction scheme:
kAB,CD
E + S 󴀕󴀬 ES → P + E (4.5)
kCD,AB

Step by step, the reaction rates are set up for each of the reaction steps involved:

RE+S,ES = k E+S,ES ⋅ cS ⋅ cE (4.6)

for the forward reaction (second-order) from enzyme and substrate to the enzyme sub-
strate complex;
RES,E+S = k ES,E+S ⋅ cES (4.7)
for the back reaction (dissociation, first-order) of the complex; and

RES,P = k ES,P ⋅ cES (4.8)

for the product formation itself.

Now we can set up material balance equations for the enzyme and the enzyme-
substrate complex. The balance equations include all reactions in which the respec-
tive molecule is converted or formed. Contributions to the balance are counted pos-
itively along the counting arrows leading out of or into the balance boundaries. The
arrows are chosen in a way that makes the numerical values positive. This gives a more
intuitive view of the final results. Rnet here means a possibly occurring accumulation.
Starting with the free enzyme the balance reads:

RE,net = −RE+S,ES + RES,E+S + RES,P = −k E+S,ES ⋅ cS ⋅ cE + k ES,E+S ⋅ cES − k ES,P ⋅ cES (4.9)
70 | 4 Kinetics – finding quantities for bioprocess reactions

For the enzyme-substrate complex we get accordingly:

RES,net = RE+S,ES − QES,E+S − QES,P = k E+S,ES ⋅ cS ⋅ cE − k ES,E+S ⋅ cES − k ES,P ⋅ cES (4.10)

This equation is the same as the one before except for the sign. Both equations are
therefore linearly dependent.
The single reaction steps are very fast, actually faster than we could observe dur-
ing the bioprocess. This gives an argument for a further simplification step: the reduc-
tion of temporal resolution. In a virtual experiment the substrate concentration shall
be kept constant by continuous feeding of new substrate. After some time we assume
that the system is in steady state meaning that the reactions take place but the observ-
able concentrations do not change. Under these conditions no accumulation of E and
ES occurs leading to:
QE,net ≈ 0 and RES,net ≈ 0 (4.11)
Furthermore, a reduction of the time horizon to reasonable periods can be envisaged,
excluding effects like enzyme aging or adding fresh enzyme. The total amount of en-
zyme protein is therefore constant and we note:

cE,tot = cE + cES (4.12)

The set of equations Equation (4.6) to Equation (4.12) now contains all a priori knowl-
edge from the mass action law and the additional simplifying assumptions. The con-
centrations of the two enzyme configurations are now obtained by solving the system
of Equations (4.11) on the left and (4.12) for cE and cES and inserting them into the bal-
ance Equations (4.6), (4.7). Note that only one of the two Equations (4.11) is necessary
and allowed, because they are linearly dependent.
cE,tot ⋅ (k ES,E+S + k ES,P )
cE = (4.13)
k E+S,ES ⋅ cS + k ES,E+S + k ES,P
k E+S,ES ⋅ cS ⋅ cE,tot
cES = (4.14)
k E+S,ES ⋅ cS + k ES,E+S + k ES,P
This calculation can be found as the Computer Algebra sheet in the supplementary
material. It allows us to deduce the required relationship between substrate turnover
and substrate concentration.
RP follows directly from Equation (4.8) by inserting Equation (4.14). Because no
metabolites are allowed to accumulate, the substrate turnover rate is directly accessi-
ble from RP :
RS ≈ RP (4.15)
Finally:
k ES,P ⋅ cS ⋅ cE,tot
RS = kES,E+S +kES,P
(4.16)
kE+S,ES + cS
This equation still leaves a certain degree of helplessness, as it is not clear from where
we get the unknown kinetic parameters without elaborating fast analytical measures.
 4.2 Enzymes as basic components – determining kinetics | 71

However, we do not need the precise value of all parameters, as they are partially lin-
early dependent and are therefore not necessarily to know separately. A common ap-
proach in process engineering is to lump several parameters into new ones, which are
relevant and comparatively easy to measure.
With
RS,max = k E+S,ES ⋅ cS ⋅ cE,tot (4.17)
and
k ES,E+S + k ES,P
kS = (4.18)
k E+S,ES
we obtain for the rate equations, which are given here on a g/L basis:

cS
R S = R S,max ⋅ (4.19)
kS + cS
as a manageable form of simple enzyme kinetics. This form of enzyme kinetics is known as Michaelis–
Menten kinetics. This is in honor of Leonor Michaelis and Maud Menten who developed the underly-
ing theory (in 1910). R S,max [g ⋅ L−1 ⋅ h−1 ] is the maximum substrate turnover rate and kS [g ⋅ L−1 ] the
limitation constant, also called the half-saturation concentration or Michaelis constant. The original
abbreviation KM is replaced here by kS to more clearly distinguish limitation constants for different
substrates. A graphical representation is given in Figure 4.3.

12
11
rS,max
10
Substrate turnover rate [g⋅L-1]

9
0.9⋅rS,max
8
7
6
5 0.5⋅rS,max
4
3
2
1 9⋅kS
kS
0
0 1 2 3 4 5 6 7 8 9 10

Susbtrate concentraon [g⋅L-1]

Fig. 4.3: Typical hyperbolic saturation curve of the Michaelis–Menten kinetics with kS = 1 g/L and
R S,max = 10 g/L. Two distinguished concentrations and the related substrate turnover rates are
marked.
72 | 4 Kinetics – finding quantities for bioprocess reactions

12
11
Substrate turnover rate Rs [g⋅L-1] 10 rS,max [g⋅g-1⋅h-1]
9 10.0
8
7 8.0
6
6.0
5
4 4.0
3
2 2.0
1
0
0 1 2 3 4 5 6 7 8 9 10
-1
Substrate concentraon cs [g⋅L ]

Fig. 4.4: Parameterized array of curves for different R S,max values.

For low concentrations substrate turnover rate reflects the equilibrium of enzyme and
substrate on the one hand and enzyme-substrate complex on the other. For the case
of cS = k S substrate turnover reaches the half maximum value as can easily be seen
by substituting k S in Equation (4.19). For high concentrations RS asymptotically ap-
proaches RS,max . However, even for nine-fold concentrations of k S only 90% of the
maximum value is reached. Here the dissociation of ES to E + P is rate limiting (Ta-
ble 4.2).
The following two figures (4.4 and 4.5) show the influence of the two formal pa-
rameters on the Michaelis–Menten curve.
To get an idea of the range of kinetic parameters some values are listed in Table 4.2.
Enzyme activity usually depends not only on substrate concentration but is also
modulated by other molecules present in cell. Their physiological meaning is to adapt
the activity of metabolic pathways to the present needs of growth conditions. In many
cases the end product of a pathway is an inhibitor, which regulates the first metabolic
step, decreasing its activity avoid overproduction. In the following paragraph the for-
mal impact of inhibitors is outlined but restricted to cases where the inhibitor binds
to the enzyme reversibly.
In the case of competitive inhibition the inhibitor molecule can bind to the en-
zyme, but the enzyme cannot convert it to a product. The situation, where the inhibitor
molecule ‘competes’ with the substrate molecule for the binding site is shown in Fig-
ure 4.6.
 4.2 Enzymes as basic components – determining kinetics | 73

12
11
Substrate turnover rate Rs [g/L]

10
9
0.5
8
1.0
7
2.0
6
4.0
5
4 8.0
3
kS [g/L]
2
1
0
0 1 2 3 4 5 6 7 8 9 10
Substrate concentraon cs [g/L]

Fig. 4.5: Parameterized array of curves for different kS values. The substrate turnover rate at a given
substrate concentration depends nonlinearly on kS .

Tab. 4.2: kCat turnover number.

Enzyme Substrate Product Organism kCat kS [mM]

Lactate dehy- L-Lactate + Pyruvate + E. coli 40 47,000 (lactate)
drogenase NAD+ NADH 180 (NAD+)
Lipase Triacylglycerol Fatty acid E. coli BL21 1500 20.4
(EC 3.1.1.3)
Invertase Saccharose α-D-Glucose + Saccharomyces 3.430 38
β-D-Fructose cerevisiae
Aminoacylase N-Acetyl-L- L-methionine + Spodoptera 134 2.72
(EC 3.5.1.14) methionine acetate frugiperda
(Sf21)
Phytase Phytate Phosphate Aspergillus 1260 0.25
niger
Penicillin Penicillin G Penicillin-Rx E. coli 0.003
amidase
Lactate dehy- L-Lactat + NAD+ Pyruvate + E. coli 40
drogenase NADH
74 | 4 Kinetics – finding quantities for bioprocess reactions

Substrate Enzyme-Substrate- Product

Complex

kE+S,ES kES,P

kES,E+S

Enzyme kE+I,EI
Enzyme
kEI,E+I
Enzyme-Inhibitor-
Inhibitor Complex

Fig. 4.6: Schematic representation of enzyme reaction mechanisms where inhibitor activity is in-
volved, in this case as competitive inhibition.

As was done in the case of the Michaelis–Menten kinetics, the formal deduction from
the single reaction constants leads to:
cS
RS = RS,max ⋅ cl
(4.20)
k S ⋅ (1 + kl ) + cS

The inhibitor only affects the apparent k S value so has a similar effect as shown in
Figure 4.5 for increasing k S . Inhibition can be overcome by increasing substrate con-
centration. The inhibition constant is chosen not as factor to cI but as a denominator
to give it the unit k I [g ⋅ L−1 ] of a concentration. This makes quantitative discussion
more practical.
There are other mechanisms of the action of the inhibitor. Unlike competitive
inhibition the inhibitor may bind to sites other than the active sites, which are called
allosteric sites. The inhibitor causes a conformational change in the enzyme affecting
reaction constants. Here different cases are distinguished. In the case of noncom-
petitive inhibition the substrate binds to the active site whether the inhibitor has
already bound or not. The inhibiting effect is caused by preventing the enzyme from
performing its catalytic action, so that only the dissociation back to enzyme and
substrate is possible. This leads to an apparent reduction of the maximum reaction
rate:
kl
k +c ⋅ c S
RS = RS,max ⋅ l l (4.21)
kS + cS
Therefore, it has a similar effect on the maximum turnover rate as shown in Figure 4.4
for decreasing RS,max . The reciprocal inhibition term could be applied to the denom-
inator with the same justification. In the chosen representation the effect may be
clearer.
 4.2 Enzymes as basic components – determining kinetics | 75

Substrate Enzyme-Substrate- Product

Complex

kS+E,ES kES,P
Inhibitor
kES,E+S

Enzyme Enzyme
kEI+S,ESI
ES+I,ESI kESI,ES+I
kE+I,EI kIS,E+I
kESI,EI+S

Enzyme-Inhibitor- Enzyme-Substrate-
Complex Inhibitor-Complex

Fig. 4.7: Schematic representation of enzyme reaction mechanisms where a generalized inhibitor
activity is involved, being described as noncompetitive inhibition.

In the case of uncompetitive inhibition the inhibitor only binds to ES but prevents ESI
from catalyzing product formation. This assumption leads to:
cS
RS = RS,max ⋅ (4.22)
k S + cS ⋅ (1 + kc ll )
This leads to a decrease in RS,max but to an apparent decrease in k S .
The three modifications (Equations 4.20, 4.21, 4.22) of the Michaelis–Menten
Equation 4.19 cover the formal possibilities of the Michaelis–Menten Equation (4.19)
with respect to numerator and two summands of the denominator. In practice the
enzymes do not always follow an either/or rule but could exhibit different mixtures
of the deduced specific cases. A generalized scheme is shown in Figure 4.7.
Inhibition is a mechanism useful for the cell to control e.g., the flux along a
metabolic pathway. In the case where the end product is the inhibitor, it prevents its
accumulation if it is not needed in smaller amounts. But the substrate itself can also
be an inhibitor when present in higher concentration. The formal description reads:
cS
RS = RS,max ⋅ (4.23)
c2
k S + cS + k l,SS

Although it looks unnecessarily complicated the inhibition constant k i,S [g/L] is writ-
ten as a denominator of the denominator to guarantee for the unit of a concentration
and to give meaningful values. For concentrations going over the limiting range sub-
strate turnover decreases again to RS,max /2 approximately at cS = k i,S ; see Figure 4.8,
where it is indicated for k i,S = 80 g/L. The optimum value defined as dRS /dcS = 0
is reached for cS,opt = √k S ⋅ k i,S at approximately RS,opt = RS,max /(1 + √k ⋅ k i,S)/k i,S ).
The meaning of substrate inhibition seems to be a bit dubious at first glance. Nev-
ertheless, in this way the cell can assign substrate, if present in excess, to another
metabolic pathway or can manage large fluctuations. In enzymatic processes and in
cultivations substrate inhibition can turn out to be a challenge for control.
76 | 4 Kinetics – finding quantities for bioprocess reactions

12
11
cS= √(ki,S⋅kS) None = ki,S →∝
Subtrate turnover rate Rs [g⋅L-1] 10
9 RS=RS,opt
8 160.0
7 80.0 cS= ki,S
6
40.0 RS= RS,max/2
5
4 20.0
3
2
-1
1 k [g⋅L ]
i,S
0
0 10 20 30 40 50 60 70 80 90 100
-1
Substrate concentration cs [g⋅L ]

Fig. 4.8: Characteristic curves for substrate inhibition; the velocity curve rises to a maximum and
then descends with increasing substrate concentration. Curves are parameterized with k i,S .

Many enzymes need a coenzyme or a second substrate to work. In these cases, the
situation becomes even more complex. In any case, no conclusions from measured
curves to the underlying mechanisms should be drawn from a process engineering
viewpoint. That is why the term ‘formal kinetics’ is often used when kinetic relation-
ships are applied to real processes.

4.3 The specific growth rate – describing growth by numbers

One of the first approaches to finding a mathematical description of growth pro-

cesses was given by the Italian mathematician Leonardo Fibonacci da Pisa (commonly
known as Fibonacci, circa 1200). He set up a sequence of numbers, namely
(n i ) = (1, 1, 2, 3, 5, 8, 13, 21, . . . ) (4.24)
where each subsequent number is the sum of the previous two. This sequence is called
the Fibonacci sequence. The idea was that a pair of rabbits grow up over one genera-
tion and then give birth to another pair. These grow up as well and become pregnant
adding their offspring after one generation time to the other rabbits making a fast in-
creasing population n i as the number of pairs grows. To the population belong the
newborn individuals from the current generation, youngsters born in the last gen-
eration being pregnant in the meantime themselves, and adults – again pregnant –
from the last generation. Note that among these there are some individuals from the
 4.3 The specific growth rate – describing growth by numbers | 77

5g/L 10g/L 20g/L 40g/L

Fig. 4.9: Photographic pictures of a growing culture in a small lab bioreactor. The tubes inside the
reactor just behind the glass wall are hardly visible even for moderate cell concentrations.

Fig. 4.10: Graphical model of the growth process; each ‘cell’ represents 5 g/L biomass concentration.

penultimate generation and so forth to the first ancestors. So individual death is not
considered in this sequence.
Before starting further calculations, the growth process can be visually followed
by a cultivation, where optical density reflects the biomass concentration as shown in
Figure 4.9. The pictures are taken in the sequence of the doubling time. During growth
the culture looks more and more turbid. The maximum value for cell dry weight is
approximately 40 g/L. Assuming 10% dry mass in the cell, the residual being water,
the cells would fill half of the reactor.
A corresponding graphical representation (Figure 4.10) shows the strongly in-
creasing cell density in a reactor.
To get an idea about how to describe the development of a growing culture it is
helpful to perform a virtual growth experiment. We start with the assumption that a
given initial number of cells e.g., nC,0 = 106 cells/ml divide after a given doubling
time td e.g., td = 1 h. This of course only holds if the cells do not react to changing
medium composition or exhibit other physiological changes like aging. The sampling
time tSample is chosen as a multiple of the doubling time td , as in this stage we do not
78 | 4 Kinetics – finding quantities for bioprocess reactions

Tab. 4.3: Data of a virtual growth experiment, where measurement of the cell number is calculated
after multiple doubling times.

Sample Sample Number of cells New cells New cells per time New cells per
number time n Cells (tSample ) ∆n Cells ∆n Cells /∆ts [h−1 ] present cells
n s [−] ts [h] [−] [−] ∆n Cells /n Cells [−]
0 0 nc,0
1 1 ⋅ td 2 ⋅ nc,0 1 ⋅ nc,0 1 ⋅ nc,0 1
2 2 ⋅ td 4 ⋅ nc,0 2 ⋅ nc,0 2 ⋅ nc,0 1
3 3 ⋅ td 8 ⋅ nc,0 4 ⋅ nc,0 4 ⋅ nc,0 1
4 4 ⋅ td 16 ⋅ nc,0 8 ⋅ nc,0 8 ⋅ nc,0 1
... ... ... ... 1
10 10 ⋅ t d 1024 ⋅ nc,0 612 ⋅ nc,0 612 ⋅ nc,0 1

have the means to calculate intermediate values. The results concerning cell numbers
nCells (tSample ) starting with the initial cell number nCells,0 = nCells (tSample = 0) of this
‘experiment’ are listed in Table 4.3 in the first three columns.
The next three columns are evaluation results. Column 4 gives the number of
newly grown cells ∆nCells as the difference between the present number of cells nCells,s
at ts and their number nCells,s−1 at the previous sampling time ts−1 . To understand
growth as a dynamical process it makes sense to relate the increase in cell number to
the observation interval ∆ts = ts − ts−1 as done in column 5. Having autocatalysis in
mind, the relation of the newly grown cells ∆nCells to the cells already present being
regarded as catalyst is sensible as shown in the last column.
Evaluation of the table results in some observations and statements.
The numbers of cells form a geometric sequence where each term after the first is
found by multiplying the previous one by a fixed, nonzero number called the common
ratio. In our case this is 2. As such it is not astonishing as we constructed the sequence
in this way. However, such geometric sequences appear quite often in describing nat-
ural or technical processes. Accumulation of these factors leads to:

ncells,s = n0 ⋅ 2s (4.25)

Not only the cell number, but also the number of new cells ∆nCells,s = nCells,s − nCells,s−1
increases in each step. The more cells present in the sample, the higher the increase
in the next time interval. The relative increase ∆ncells,s /nCells,s−1 is therefore always
constant (last column). This last statement represented as mathematical expression
reads:
∆ncells ∼ nCells,s−1 (4.26)
 4.3 The specific growth rate – describing growth by numbers | 79

To understand growth as a dynamic process it makes sense to relate the increase dur-
ing a time interval to this time interval as:
∆nCells
∼ nCells,s−1 (4.27)
∆tsample

This expression is not a satisfactory result for a dynamic process as an interval is re-
lated to its lower boundary. Practically speaking, if somebody measured the cell num-
ber only every two hours and interpolated in between, they would get for the time
interval between sample number 2 and 4 a value for nCells of (16 − 4)/2 = 6 ⋅ nCells,0
instead of 8 ⋅ nCells,0 . Consequently, a transition to infinite intervals is necessary:

dnCells
∼ nCells (4.28)
dt
Not all cells will divide at the same time and are not of equal size. To generalize this
observation the cell number can be substituted by the cell dry mass m X :

dm X (t)
∼ μ ⋅ m X (t) (4.29)
dt
Now it is reasonable to introduce a proportionality factor μ:

dm X (t)
= μ ⋅ m X (t) (4.30)
dt
The proportionality factor μ is called the ‘specific growth rate’. The term ‘growth’ rep-
resents the newly formed biomass, ‘rate’ means per time and ‘specific’ per biomass
present. The unit is subsequently μ [g ⋅ g−1 ⋅ h−1 ] or shorter [h−1 ]. Finally, the defini-
tion for the specific growth rate is:

dm X (t)
μ= (4.31)
m X (t) ⋅ dt

With help of the growth Equation (4.31) we can now calculate the course of cell dry
weight during a cultivation. This differential equation can be solved by separation of
the variables
dm X (t)
= μ ⋅ dt (4.32)
m X (t)
and integration of both sides separately:

dm X (t)
= μ ⋅ dt (4.33)
m X (t)
m X,t t
dm X (t)
∫ = ∫ μ ⋅ dt (4.34)
m X (t)
m X,0 t0
80 | 4 Kinetics – finding quantities for bioprocess reactions

In mathematical textbooks the general primitive of a function of the form f(x(t)) =

dx(t)/x(t) is given as F(x) = ln(x(t)). Applying this to our problem results in:

m X (t)
[ln m X ] = [μ]0t → ln m X (t) − ln m X,0 = μ(t − 0) → ln = μt (4.35)
m X,0

Note that the integration of the right side is only possible for constant μ. Application
of the exponential operator finally gives the required growth equation:

m X (t cult) = m X,0 ⋅ eμ⋅tcult (4.36)

This exponential function is the most fundamental equation to describe growth pro-
cesses. It is also used in other fields in cases where the state of a system is proportional
to its changing rate. However, it is valid only for growth processes with constant μ. The
Fibonacci sequence can be historically interpreted as a rough approximation of an in-
tegration for the case of natural numbers.
The course of cell dry mass concentration of an exponentially growing culture is
shown in Figure 4.11. It is impressive that the curve increases faster and faster from the
small value of 1 g/L to 100 g/L after less than seven doubling times.
Not all of the different classes of microorganisms can grow so fast even under op-
timum conditions. In their natural habitat growth is limited by availability of nutrition
or other environmental influences. Furthermore, the population is subject to cell death
by many means. In bioreactors growth rates can be kept close to optimum for only a
few generations. In many cases there are reasons to keep growth rates lower via an
intentionally applied substrate limitation. The relation between substrate supply and
growth is outlined in the next subsection; application examples are given in the next
chapters (Figure 4.12).
Typical spans of specific growth rates of bacteria range from 1.4 h−1 at maxi-
mum to 0.1 h−1 during production. Many eukaryotes like yeasts and filamentous fungi
can reach rates of 0.5 h−1 but are often cultivated at lower rates. Microalgae and sus-
pended plant cells are much lower in the range of 0.03 h−1 ≈ 0.69 day-1 , which cor-
responds to a doubling time of t d = 1 day. Animal cell cultures often reach doubling
times of only one week. The upper limits are only valid for optimum conditions and
may vary strongly from species to species.
 4.3 The specific growth rate – describing growth by numbers | 81

120
110
100 c X(t) = c X,0 ⋅exp(μ⋅t)
Cell dry weight [g g- h-1]

90
80 c X,0 = 1 g/L μ = 0,25⋅h-1
70 64 = 1⋅26
60
50 e4 = 54⋅6
40
30
20
10
4/μ= 16 h 6⋅td=16⋅6 h
0
0 10 20
Culvaon me [h]

Fig. 4.11: Cell dry weight of an exponentially grown culture with the constant specific growth rate
μ = 0.25 h−1 .

120
110
100
Cell dry weight cx [g g-1h-1]

90
80
1.0
70
0.8
60 μ [h-1]
0.6
50
0.4
40
30
0.2
20
10
0
0 10 20
Culvaon me tc [h]

Fig. 4.12: Cell dry weight of exponentially grown cultures with different specific growth rates.
82 | 4 Kinetics – finding quantities for bioprocess reactions

4.4 The yield coefficient – combining substrate turnover

and growth

The question we are grappling with in this paragraph is how the cells ‘know’ how fast
they can grow. Many environmental parameters like temperature, pH, or salinity of
course play a role. Intracellular enzymatic steps also influence possible growth rates
differently for different groups of organisms. Even in the case that all environmental
parameters are at their respective optimum the cell is limited by intracellular steps
e.g., if an enzyme is at its maximum turnover rate. This gives reason to define a maxi-
mum specific growth rate μmax [g/g/h]. It is the task of metabolic modeling to find out
the intracellular reasons for the value of μmax for different classes of microorganisms.
In the next paragraphs we focus on the role of substrate concentration.
From previous experience it is known that substrate concentration is one ma-
jor factor that can be influenced by process engineering means. In the discussion of
medium design (Chapter 3) we already used the relationship between initial substrate
and final biomass concentration. From the experiment (Figure 4.1) it becomes obvi-
ous that with progressively increasing biomass concentration substrate concentration
also decreases faster, so we assume that growth and substrate uptake are coupled. In
the time interval where cS approaches zero growth is no longer possible, implying that
growth depends on the actual substrate concentration.
The first attempt to find a more or less unspecific relation between growth and
a limited environmental resource was formulated by Pierre-François Verhulst (1854).
The background was the observation that growth, e.g., as predicted by Fibonacci, can-
not go ad infinitum but has to stop somewhere. The imagination was a limited area
offering enough space or food for a couple of animals but with a limited carrying ca-
pacity K of individuals. As this capacity becomes exhausted the animals have to mi-
grate or do not get so many offspring due to starvation. Food like grass could further
grow but is divided by more and more animals. Finally, each individual gets so little
food that only survival but no propagation is possible. In addition to the linear growth
model (4.29) Verhulst postulated that growth depends on the residual capacity (K-N)/K
(normalized from 0 to 1) of the habitat:
dN (K − N)
= μ max ⋅ N ⋅ (4.37)
dt K
The solution of this differential equation with N0 being the initial population is:

K ⋅ N0 ⋅ eμmax ⋅t
N (t) = (4.38)
K + N0 ⋅ (eμmax ⋅t − 1)

This so called ‘logistic growth curve’ is a sigmoid function and a popular (as it has at
least a closed solution) first trial to fit growth data. Nevertheless, the assumptions do
not really represent the situation in a bioreactor. In our batch, for example, we see that
growth ceases only very late, when glucose is nearly exhausted.
 4.4 The yield coefficient – combining substrate turnover and growth | 83

rS

μ = rX

rP

Fig. 4.13: The concept of substrate limitation; an enzymatic transport step (symbolized by the turn-
ing enzyme) carries substrate into the cell. This substrate flux determines the possible growth.

The first one to describe growth as a function of substrate concentration was Jacques
Monod in 1942 when he gave the famous Monod equation:

cS
μ = μmax ⋅ (4.39)
kS + cS

Monod won the Nobel Prize (Physiology or Medicine) in 1965 for the operon theory, not
for this equation. Since then the Monod equation has been employed to describe myr-
iad bioprocesses. The equation shows structural similarity to enzyme kinetics, so the
assumption has been formulated that enzyme kinetics is an underlying mechanism.
Substrate uptake as an enzymatic step in metabolism is the first candidate to investi-
gate as shown in Figure 4.13. Understanding that the more biomass is present the more
substrate will be taken up, it is sensible to define the specific substrate uptake rate rS
in analogy to the specific growth rate:
dmS
rS = (4.40)
dt ⋅ m X
as a measurable property of the cells. Following the attempt to understand the Monod
equation and in analogy to enzyme kinetics rS can be given as a function of the sub-
strate concentration cS :
cS
rS = rS,max ⋅ (4.41)
kS + cS
This is in accordance with the experiment in Figure 4.1 that at lower substrate concen-
trations the substrate uptake rate is reduced. Substrate availability inside the cell then
determines the possible growth rate. This concept is called ‘substrate limited growth’.
The substrate is then converted by hundreds of enzymes inside the cell. However,
growth is regarded as a thermodynamic process where the cell could make ‘the best
of it’ under the restrictions of energy and material balances, stoichiometric relations,
and other constraints. We have already discussed the elemental balance as such a con-
straint. Biology has provided different approaches to understand how much biomass
84 | 4 Kinetics – finding quantities for bioprocess reactions

can be built up by a given amount of substrate taken up. Of specific interest is the
partitioning into anabolic pathways lumped as growth on the one hand and energy
producing metabolic pathways on the other. A fundamental relationship determining
the relation of growth and substrate uptake was found by S. John Pirt in 1965:
1
rS = ⋅ rX + rm (4.42)
y X,S

To get a consistent system of terms the historic symbol μ is replaced by r X , where r

is rate and X is biomass. To maintain a given specific growth rate the cell has to take
up a proportional rate of substrate. An important point is that the proportional factor,
interpreted as yield coefficient, is constant for many growth conditions. This has to be
understood as an optimized carbon allotment. Metabolic energy is provided by a part
of the substrate while the other part is used for growth under energy utilization. This
balance is kept stable by the cells in order to avoid waste of energy and carbon. As
long as the cell does not change this metabolic pattern this results in a constant yield.
In addition, energy is partly used for nongrowth-associated purposes like movement,
keeping the water balance, or maintaining membrane potential. This part of energy
utilization is lumped into the ‘maintenance’ term r m (m in the original reference). In-
deed, very often measurement data fit quite well to the linear relation between growth
and substrate uptake. For clarity it is noted that y X,S = dr X /drS is a kinetic parameter,
while Y X,S = ∆m X /∆mS is a measured process characteristic over a longer cultivation
period. About the inner connection between substrate uptake rate and growth rate
more background is given in the next chapter.
A first idea to describe biomass formation and substrate uptake simultaneously is
simply combining the Monod and the Pirt equation ((4.33) and (4.37)). A short inspec-
tion of this system of equations suggests that it is not a good idea. For cS = 0 a positive
substrate uptake rate is predicted, which is physically not possible. Turning the cause-
effect chain we set the substrate uptake as a primary limiting step and reformulate the
resulting growth rate following Pirt’s linear relationship:

rS,max ⋅ c S
rS = (4.43)
kS + cS
r X = y X,S ⋅ rS − r m (4.44)

This is a first simple physiological model of microbial growth and will be used as an
example throughout this book. According to the essence of kinetics these equations
hold irrespective of the way cS changes during the process.
Besides biomass formation and substrate consumption, product formation is of
interest. As was done before we define a specific product formation rate:

dmP
rP = (4.45)
dt ⋅ m X
 4.4 The yield coefficient – combining substrate turnover and growth | 85

In addition, a formal relation with growth and/or substrate turnover is required. As

the cell is highly stoichiometrically determined again a linear relation in analogy to
Pirt’s equation can be a good start:

rP = yP,X ⋅ r X + k P (4.46)

This equation was introduced to the field by Luedeking and Piret and bears their
names. It is a flexible tool to fit different kinds of product formation patterns. The yield
coefficient yP,X = drP /dr X describes the ratio of product formed per biomass growth.
Often product formation is coupled with catabolism, meaning substrate turnover. At
least the carbon comes from there. For technical applications the ratio of product
produced to substrate consumed yP,S = ∆mP /∆mS is of interest. This is considered in
the reformulation:
rP = yP,S ⋅ rS − rP,m (4.47)
Here product formation is coupled to substrate uptake where a small constant part of
the substrate is reserved for other purposes. This holds only for rS and r X over a given
threshold, which is also the case for the Luedeking–Pirt equation. In the next chapter
and some case studies the coupling of product formation to catabolism and anabolism
(rS , r X ) and the complexity of control patterns of the cells is analyzed in more detail.
Before trying out the kinetic equations for our standard batch process good
guesses for reasonable values of the kinetic parameters are required. Table 4.4 gives
some typical values.
Some trends can be observed from Table 4.4. Aerobic microorganisms can reach a
yield of 0.5 g/g growing on glucose. Sometimes it is a bit lower, but never higher. The
yield also depends on the substrate. With a higher degree of reduction (ethanol) higher
yields are obtained. Anaerobic growth on glucose results in a low yield of about 0.03–
0.05 g/g. Here, the product is always formed in a high yield product per substrate of
e.g., 0.5 g/g. As substrate uptake is often higher for anaerobic growth, similar growth
rates as for aerobic growth are obtained.

Tab. 4.4: Typical specific growth rates of different microorganisms and substrates.

Organisms Substrate r S,max r X,max kS y X,S Remarks

[g/(g ⋅ h)] [g/(g ⋅ h)] [g/L] [g/g]
Sacch. cerev. Glucose 2.52 0.44–0.60 0.180 0.485 Aerobic
Ethanol 0.12–0.14 0.753 Anaerobic
Glucose 0.53–0.63
E. coli Glucose 0.5–1.0 0.004 0.5 Aerobic
Aspergillus oryzae Glucose 0.005 Aerobic
Klebsiella aerogenes Glucose 0.009 Aerobic
Glycerol 0.009
Virt. generic Glucose 1.0 1.0 0.5 Aerobic
Virt. generic Glucose 1.0 1.0 0.05 Anaerobic
86 | 4 Kinetics – finding quantities for bioprocess reactions

4.5 The batch process – the simplest form of a bioprocess

Now all the information can be assembled to get a simple model to describe a batch
process. Remembering the physiological model
rS,max ⋅ cS
rS = (4.48)
kS + cS
r X = y X,S ⋅ rS − r m (4.49)
rP = yP,S ⋅ rS − rP,m (4.50)

the concentrations c X and cS have to be calculated to have a complete set of equations.

This has to be done by setting up the material balances for biomass and substrate. For
the batch process the balance equations are in the form of simple differential equa-
tions of accumulation and reaction terms:
dc X
= μ ⋅ cX (4.51)
dt
dcS
= −rS ⋅ c X (4.52)
dt
dcP
= rP ⋅ c X (4.53)
dt
These reactor equations together with the physiological equations form a complete
model to represent a batch cultivation. Note the minus sign for substrate to get a
decreasing substrate concentration for a positive turnover rate. Even for more com-
plicated cases the biological submodel and the reactor submodel should always be
strictly separated during building up and for writing them down. Unlike in setting
formal kinetics, e.g., only for biomass, these models cannot be solved explicitly but
must be numerically integrated. A simulation is shown in Figure 4.14.
Batch processes are usually divided into several phases. During the lag phase the
cells have to adapt to the environmental conditions inside the reactor, which may be
different from the parameters in the preculture. The specific growth rate increases only
after some time, e.g., one hour, to the value predicted by kinetics. This can happen e.g.,
in cases where the medium in the preculture contains supplementary compounds that
are not present in the technical medium or vice versa. Adaptation is not represented
in the model equations. As the lag phase is unproductive it should be kept as short as
possible e.g., by adjusting the conditions in the shaking flasks or the cultivation in the
stage before.
During the ‘exponential phase’, μ is close to but not equal to μmax for several hours
as substrate concentration is far in the saturation range of Michaelis–Menten kinetics.
Do not use the outdated term ‘logarithmic phase’. This is actually the state of the reac-
tor we want to achieve as long as possible. This is no longer the case in the ‘stationary
phase’, where substrate concentration falls to the order of magnitude of the limitation
constant and becomes therefore the decisive value for growth. In many cultivations ac-
cumulation of an (possibly unknown) inhibiting product could also contribute to the
 4.5 The batch process – the simplest form of a bioprocess | 87

120
110

staonary phase
0.4
100

decay phase
cX [gL-1]
cS [gL-1]
90 Specific rates
Specific rates 0.3
80
cX [gL-1], cS [gL-1]

Specific rates
70
0.2
60 exponenal phase
50
lag phase

0.1
40
30 0.0
20
10 -0.1

0
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
Culvaon me [h]

Fig. 4.14: Simulation of a batch cultivation with the parameter set for the model organism Virtuella
generica (Table 4.4) showing the course of the concentrations and the specific turnover rates.

stationary phase. After the substrate is completely exhausted a decrease of cell dry
mass is observed, which is also visible by the negative specific growth rate. That does
not mean that the cells really perish in the ‘decay phase’. In fact, they fuel their main-
tenance demand by respiration on intracellular materials like storage compounds, a
process called ‘endogenous respiration’. The optimum harvesting point is shortly be-
fore entering the decay phase.
We started this chapter by evaluating a batch process in terms of macroscopically
observable volumetric rates R and yields Y (capital letter). These characteristic values
have to be understood as integral values over a longer time interval. Enzyme kinetics
could be deduced from the mass action law being a basic principle in chemical re-
actions. Starting from the biological observation of propagation by cell division and
doubling, the specific growth rate r X was introduced. Like the other specific rates r this
is understood as a differential value being valid at all time instances. The same holds
for the model parameters k and y (lowercase letter). Kinetics in combination with ma-
terial balances are the key to formulate descriptions of a growth process with at least
some mechanistic justification. The basic assumption was the translation of enzyme
kinetics to growth kinetics. In the next chapters we will see how far we can go with
this daring idea.
88 | 4 Kinetics – finding quantities for bioprocess reactions

4.6 Exercises, questions and suggestions

1. What is the relationship between the specific growth rate and the doubling
time?
2. In 2010 the specific growth rate of mankind was 1.2% per year. After how
many years is the number of people expected to have doubled?
3. Some bacteria have a specific substrate uptake rate of 10 g ⋅ g−1 ⋅ h−1 . If a hu-
man had the same metabolic speed, how many sacks of sugar would they
need per day? Discuss the result compared to the approximated real energy
need of a human. How many orders of magnitude would it be more or less?
4. Evaluate the batch process of Figure 4.1 and Figure 4.14. Instead of using dif-
ferences as in Table 4.1 use the exponential growth formula (4.36) to calculate
μ for the given time intervals.
5. Discuss the results from Table 4.1 and from Exercise 4.4.
6. Recombinant proteins shall be produced in a batch culture. The specific
growth rate is μ + = 0.2 h−1 . In the inoculum 1% of the cells have lost their
plasmid and can now grow faster with μ− = 0.3 h−1 because they do not
spend metabolic energy on product formation. What is the percentage of cells
without the plasmid after tEnd = 12 h?

1. An initial number of cells m X,0 doubles after t d to

1
m X,0 (t d ) = m X,0 ⋅ exp(μ ⋅ t d ) = 2 ⋅ m X,0 → exp(μ ⋅ t d ) = 2 → μ = ⋅ ln(2) (4.54)
td
2.
1
td = ⋅ ln(2) → t d = 57.8 years (4.55)
μ
3. Additional given values: mass of one sack of sugar mSack = 25 kg; recommend
daily energy requirement of human = 10,467 MJ with a mass of 62.5 kg (≈ to 25 kg
dry weight); energy content of sugar per kg = 16,957 MJ.
Mass of sugar needed:

kg kg
mS = rS ⋅ m X,human = 10 ⋅ 24 ⋅ = 6000 (4.56)
kg ⋅ d ⋅ 25 kgX d

Number of sacks sugar nSacks is equal to:

mS 6000 kg
d 1
nS = = = 240 (4.57)
mSack 25 kg d

A human with the same specific substrate uptake rate would need 240 sacks of
sugar per day.
 4.6 Exercises, questions and suggestions | 89

Food needed for metabolic energy for humans calculated as sugar:

MJ
Energyhuman 10.467 d kg
mS = = MJ
= 0.617 (4.58)
Energysugar 16.957 kg d

Calculated order of magnitude:

mS,Human=Bacteria 6000 kg
d
ratiobacteria,human = = = 9720 (4.59)
mS,realHuman 0.617 kg
d

We conclude that under these reference conditions, bacterial consumption of en-

ergy per unit biomass is about four orders of magnitude higher than that of a hu-
man.
4. To evaluate data from discrete measurements some assumptions are necessary.
The simplest one is that μ is the remaining constant between two measurement
points. This neglects time changing effects of limitation and maintenance. The
exponential growth formula can then be used for evaluation of μ:

1 c X (t2 )
c X (t2 ) = c X (t1 ) ⋅ eμ⋅(t 2 −t 1) → μ = ⋅ ln ( ) (4.60)
t2 − t1 c X (t1 )

or in a simpler notation:
1 cX2
μ= ⋅ ln ( ) (4.61)
∆t cX1
Now we assume that y X,S and rS are constant as well:

∆c X μ
y X,S = ; rS = (4.62)
∆cS y X,S

It becomes obvious that the solution of the differential equation is more accurate
than using differences. For very low substrate concentrations unrealistic values
for rS are observed. This is due to respiration on intracellular storage compounds
making the simplified assumption rS = μ/y X,S wrong for negative μ, as then no
substrate is produced. Extreme care thus has to be taken when applying formal
equations without knowing the background and constraints.
To determine the parameter y X,S and μe for the whole physiological model we
plot μ versus rS (Figure 4.15). We see that our basic physiological model holds.
The slope of the straight line is the yield coefficient and the intercept at the μ axis
the maintenance. The limitation constant k S cannot be determined from batch
processes as the limited phase is very short.
Remark: To plot a growth curve on logarithmic paper and make a linear regression
gives in principle a value of μ. Nevertheless, it is not up to date in the computer
age. The same holds for the Lineweaver–Burk diagram, a transformation of en-
zyme kinetics. Calculating the reciprocal Michaelis–Menten kinetics a linear plot
is obtained, which allows us to read the kinetic parameters. Measurement errors
90 | 4 Kinetics – finding quantities for bioprocess reactions

Tab. 4.5: Calculated values from the measurements in Figure 4.1.

t [h] c X [g/L] c S [g/L] μ [1/h] y X,S [g/g] r S [1/h]

0.0 5.0 100.0
−0.0454 −0.1667 0.2721
0.9 4.8 98.8
0.1332 0.3552 0.3749
2.7 6.1 95.1
0.2774 0.4066 0.6823
5.4 12.9 78.4
0.2393 0.4729 0.5060
6.3 16.0 71.9
0.2419 0.7139 0.3388
8.1 24.7 59.6
0.2330 0.3589 0.6492
9.0 30.5 43.6
0.1950 0.4554 0.4283
9.9 36.4 30.7
0.1495 0.4674 0.3198
10.8 41.6 19.5
0.1343 0.3406 0.3944
11.7 46.9 3.8
0.0183 0.2103 0.0870
12.6 47.7 0.1
−0.1232 −4.65E+01 2.65E−03
13.5 42.7 1.6E−03
0.0248 1.22E+03 2.02E−05
15.3 44.6 4.3E−07

are also transformed leading to limited accuracy. This gives a first idea about an
appropriate model selection, but later computer evaluation is more reliable.
To be sure about the choice of y X,S and rS here the formal deduction is given:
Solving the balance equation for substrate at an interval ∆t and substituting the
exponential growth equation leads to:

d rS ⋅ c X (t1 ) ⋅ eμ⋅t
cS (t) = −rS ⋅ c X (t) = − (4.63)
dt eμ⋅t 1
rS ⋅ m X (t1 ) ⋅ eμ⋅(t−t 1) rS ⋅ c X (t1 )
cS (t) = − + cS (t1 ) + (4.64)
μ μ
Solving for rS and rearranging with t=t2 leads to:
μ ⋅ (cS (t) − cS (t1 )) μ ⋅ ∆cS ∆cS 1
rS = − =− = −μ ⋅ = μ⋅ (4.65)
c X (t1 ) ⋅ (e μ⋅(t−t 1 ) − 1) c X (t1 ) ⋅ (e μ⋅∆t − 1) −∆c X y X,S

This justifies our initial assumptions to take y X,S directly from the data and calcu-
late rS from μ and y X,S .
 4.6 Exercises, questions and suggestions | 91

0.4
value error
Intercept -0.04083 0.03799
Slope 0.48527 0.09495
0.3
Chi-Quadr 0.00496
R-Quadrat 0.72313
Pearson R 0.85037
0.2
µ [1/h]

0.1

0.0

-0.1

-0.1 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
rS [1/h]

Fig. 4.15: Estimation of μ and rS from data (Figure 4.1) and calculation of y X,S from the regression
curve.

5. Volumetric productivity is low at the beginning because of low biomass concen-

trations.
The formal parameter y X,S is 0.5 g/g in the model, but the observable macroscopic
Y X,S is lower due to lag phase and maintenance. μ slowly becomes lower during
cultivation because cS approaches the order of magnitude of k S and maintenance
becomes more and more dominant.
6.
m X,0,− = 0.01 ⋅ m X,0,+ ;
m X,12,− (4.66)
= 0.01 ⋅ exp(μ − − μ + ) ⋅ tEnd ≈ 0.033 = 3, 3%
m X,12,+
The subpopulation that lost the plasmid has an evolutionary advantage in the
reactor.
5 Bioreactors – designing a home for the bioreaction
Bioreactors show a huge diversity depending on the product, the microorganisms em-
ployed, and the production environment. A bioreactor cannot be simply bought off
the peg. This chapter starts by considering what the different purposes are. This is
also the starting point for a biochemical engineer to design new reactor concepts with
new features. In order not to reinvent the wheel a look at commonly applied reactors,
taking on board the terms and concepts to describe their performance is helpful. For
this starting point the stirred tank reactor is chosen. From there the view goes to dif-
ferent reactor types specialized for bioprocesses. This includes, in the last paragraph,
a walk through several orders of magnitude of dimension to assess current designs.

5.1 Not only a vessel – for what a bioreactor is needed and what it
can be

On hearing the word ‘bioreactor’ many people think immediately of a stainless steel
vessel or a small glass reactor in the lab. Here we want to avoid this shortcut thinking
and try to specify the requirements applying to the reactor. This includes how the reac-
tor supports the goals of the process and the technical constraints. In this book there is
no space to set up a complete ‘user requirement specification’ (see supplementary ma-
terial) but is important to realize the basic concepts. First of all the bioprocess should
take place in a defined space, where the biocatalysts, microorganisms, or enzymes do
their job without being dispersed into the environment. This it is something like an
exoskeleton, to stress a picture from biology. Secondly, to enable the microorganisms
to grow and carry out material conversion, suitable conditions with respect to tem-
perature, pH or medium concentration have to be maintained. This implies, thirdly,
transport of water and chemical compounds into and out of the reactor. With these
very general statements a general definition can be given:

“A bioreactor is an enclosure in which, in the presence and involvement of a biocatalyst material,

transformations take place.”

This definition does not imply that the reaction space is closed in a strict sense. De-
pending on the specific task, different exchange mechanisms are at work but have to
be defined and controlled. On the other hand, the definition implies that the reactor
is a manufactured device. These last two points make up the difference e.g., to the
biofilm on a stone or to our intestine, where material exchange also happens and mi-
croorganisms feel happy. In terms of process engineering a bioreactor is a technical
transport/reaction system. Starting from an understanding of bioprocesses we now
collecting things to be transported.

https://doi.org/10.1515/9783110315394-005
 5.1 Not only a vessel – for what a bioreactor is needed and what it can be | 93

Manometer

Foam
An-
Alkali
Inoculum culture
Cooling trap
and nutrient Filter Exhaust air
soluon

Foam sensor
Steam Cooling water

pH-sensor Measurement
Cooling jacket
Srrer and control
technology
Baffle Thermal sensor

Cooling
water Steam
Sprinkler

Air Filter
Harvest
control
Engine

Fig. 5.1: Scheme of a typical bioreactor with peripheral devices. Follow all transport lines into and
out of the reactor while reading the next paragraphs.

The medium has to be pumped from a medium flask or container through pipes into
the reactor. The included substrates have to be further transported via the next con-
vective step all through the reactor space to the cells. This usually involves mixing.
The last step is diffusion to and enzymatic transport by the cells. Other liquids to be
dosed are the inoculating material, and titration and antifoam agents. Following the
process further we come to sampling and harvesting. Similarly, gaseous substrates
enter the reactor driven by pressure, and are then macroscopically spread by bubbles
and further by mixing and diffusion. These processes require peripheral equipment
outside the reactor like pumps, valves and pipes, and specific structures inside like
stirrers. Figure 5.1 gives an overview. More details are given below and in the chapter
on fermentation (Chapter 11). Energy also has to be provided and to be removed. Me-
chanical energy is needed to drive most of the transport processes. It is transported via
the stirrer shaft and dissipated by the impeller. In fact, this is a critical point in reactor
design and needs special attention. All bioprocesses are in principle exothermic, so
we have the issue of transporting heat out of the reactor. All material and energy fluxes
have to be somehow observed and controlled, so transport of information is an indis-
pensable issue. Different sensors make process variables accessible, while actuators
control fluxes according to the needs of the microorganisms and according to our idea
of the process. More about measurement techniques is given in Chapter 10. Unfortu-
nately, gathering information is sometimes neglected, making process development a
game of trial and error.
94 | 5 Bioreactors – designing a home for the bioreaction

The needs of microorganisms are at the center of the considerations. But outside
the reactor humans work with their own requirements, so we have to respect these
as well. ‘Safety first’ is the motto of labor and environmental protection. No microor-
ganisms should leave the reactor unintentionally e.g., via the off-gas. Neither should
any microorganisms enter the reactor to keep an axenic culture, which is important
for product safety and economic risk minimization. Of course the whole equipment
is subject to governmental regulations. Specific features of bioreactors that address
these issues are summarized under the terms ‘aseptic’ or ‘hygienic design’, for exam-
ple heat (steam) sterilization before starting the cultivation. Small reactors have to
be brought to an autoclave, while large reactors have their in-built devices that allow
sterilization in place (SIP). Cleaning is also an issue; a typical housewife and house-
husband problem. Nobody likes to scrub a fermenter, as it is sometimes called disre-
spectfully. Large reactors would need to be entered by laborers using mountaineering
equipment. In modern reactors appropriate installations make the reactor capable of
cleaning in place (CIP). Manual operation is also not state of the art but automation
based on the sensors and actuators mentioned above is.
Economic efficiency is the item a company has to look for. This has impact on the
material, number of sensors or (CAPEX, OPEX) and energy savings. Bioreactors are
not only used for production but also for scientific reasons. Highly equipped biore-
actors have advantages over shaking flasks, which are principle bioreactors as well.
This concerns the possibility to measure and control environmental variables, allow-
ing for frequent sampling or delivering of more reproducible results. Typical biomass
concentrations are tenfold higher due to better oxygen transfer. Here the costs are not
balanced against a chemical compound as valuable product but against amount of
data and quality of gathered information. This justifies the costs of better equipment.

5.2 Revisiting the stirred tank reactor – the most important

issues to learn from chemical engineering

Stirred tank reactors are the most widely used reactors in chemical and biological pro-
cesses due to advantages such as simplicity of construction, low capital and operat-
ing costs, and operational experience. A stirred tank reactor is a cylindrical vessel in
which the uniform mixing of the reactants is achieved by means of an integrated stir-
rer or agitator. Since a bioreactor is the place where the biochemical reaction takes
place, bioreactors are the heart of the bioprocesses.
Standard geometry of bioreactors may vary depending on the process and or-
ganism requirements. The reactor shape is mostly cylindrical with a flat bottom, but
curved, dished and conical bottoms are also used. Curved bottoms support the mixing
of the medium especially at the bottom and prevents sedimentation at the edges. How-
ever, a standard design with a flat bottom is preferable due to the operational costs.
5.2 Revisiting the stirred tank reactor – most important issues from chemical engineering | 95

The bioreactor construction material is typically made of glass if small scale bioreac-
tors are needed. For industrial applications and pilot scale implementations, stainless
steel installations are used as construction material. High quality (Cr-Ni, V4A) is nec-
essary to prevent corrosion induced by high salt contents and nickel discharge. Elec-
trolytic polishing of the surface to a finish roughness of 0.8 µm is preferred to prevent
microorganisms attaching to small grooves. The ratio of height (HR ) to diameter (DR ),
referred to as aspect ratio, of a stirred tank bioreactor is usually chosen between 2–3
depending on the number and arrangement of impellers and the reactor application.
A higher aspect ratio enables better aeration, heat transfer, and mixing as it is used
for microbial applications. Lower aspect ratios are generally needed when animal cell
cultures are used since reduced shear stress, mixing, and aeration is essential. There
are three basic reasons for a vertical shaped bioreactor: The contact time between bub-
bles and liquid is longer and therefore oxygen can be better treated in the bioreactor.
The reactor can be better cooled down due to the higher specific area (area/volume).
Better lighting area is achieved for phototrophic organisms
Adequate mixing and aeration are important tasks for bioreactors. Agitation for
the purpose of mixing is obtained by the power driven agitators and the baffles. Baf-
fles are used to increase turbulence and prevent vortexing and swirling of the fluid.
The number of baffles ranges between four and six depending on the diameter of the
bioreactor. Dimensions of baffles and impellers are a fraction of the reactor diameter
and are depicted for a typical case in Figure 5.2.
The use of multiple impellers improves mixing and effective gas-liquid mass trans-
fer. The common working volume of a bioreactor is 70–80% of the total volume of the
bioreactor. The remaining volume, i.e., the headspace, is a restraint system for foam
and depends on its formation. Furthermore, increasing aeration rate brings more bub-
bles into the system and the liquid surface needs space to rise. The same holds for the
liquid rising at the reactor wall due to the central vortex. Gas is introduced into the
bioreactor via a perforated sparger, which is located under the bottom impeller, with
a moderately smaller diameter compared with the impeller diameter. Besides the mi-
nor influence of the sparger, impeller type plays an important role in terms of gas dis-
persion. A wide variety of impeller sizes and shapes allows us to create various flow
patterns inside the bioreactor. Impeller type is basically chosen depending on the vis-

D/12- D/10
D/72 - D/50
2D - 3D

D/3 - D/2

D/3 D/3 Fig. 5.2: Geometry of a typical stirred tank reactor; most of the
D dimensions are related to the reactor diameter.
96 | 5 Bioreactors – designing a home for the bioreaction

Rushton Turbine Propeller Intermig-Agitator

Fig. 5.3: The three most important impeller types for bioreactors.

Gas outlet Gas outlet

Dra tube

transport
Axial

Gas inlet Gas inlet

Srred tank reactor (STR) Dra tube reactor
Engine Engine
(a) (b)

Fig. 5.4: The basic flow patterns of the Rushton turbine (a) and the propeller (b), being different in
primarily generating radial or axial flow respectively.

cosity of the fluid and the sensitivity to shear stress. Three different impeller types are
commonly applied in stirred tank bioreactors. These impellers are three-blade pro-
peller, six-blade Rushton turbine and Intermig (Figure 5.3).
For flat-bladed turbines the most common generic term is ‘Rushton turbine’ af-
ter to J. H. Rushton. Their blades are flat and set vertically along an agitation shaft,
which produces a unidirectional radial flow; compare in Figure 5.4. The standard de-
sign features six vertical blades, although four and eight are also common. Rushton
type impellers are commonly used in fermentations of cell lines that are not consid-
ered shear-sensitive, including yeasts, bacteria, and some fungi. Propeller type agi-
tators resemble the design of marine propellers, in that case to generate propulsion
for ships. Consequently, propellers transfer water mainly in the axial direction. To-
gether with bending of the flow at the bottom of the reactor a low shear axial mixing
is achieved. The industrially developed Intermig agitator is a predominantly axial-
5.2 Revisiting the stirred tank reactor – most important issues from chemical engineering | 97

pumping impeller that comprises several stages arranged on top of one another and
each staggered by 90 °C angles. A typical layout has two stages and two flow direct-
ing elements. The decisive difference between the Intermig and other mixing systems
is the arrangement of the two blades: the inner blade and the two-stage outer blade
are oriented in the same radial plane but with opposing blade angles. This generates
a primary axial pumping flow in opposing directions in the inner and outer sections
therefore also supporting axial mixing.
Rotation of stirrers in a stirred tank bioreactor is enabled by application of electri-
cal power. From the agitator mechanical energy flow goes further into the suspension.
Here the specific (volumetric) power input drives hydrodynamic behavior determining
mixing, and heat and gas transfer, the most important processes for operation and
scale up. So, first, the power input PR,imp [W] has to be characterized. Power input
depends on the rotational speed of the stirrer, geometry of the stirrer and fluid charac-
teristics. Basically it can be understood as a product of torque, being the driving force
of the blades multiplied with the lever arm, and the angular velocity. This results in
Equation (5.1):
3
PR,imp = NP ⋅ ρ liqu ⋅ Nimp ⋅ D5imp (5.1)
PR,imp is the transferred power, ρ liqu [kg ⋅ m−3 ] is the fluid density, Nimp [s−1 ] (in prac-
tice often given in min−1 ) is the rotational speed of the stirrer, and Dimp is the diameter
of the impeller. There are some uncertainties with respect to this expression, e.g., con-
cerning the force on the blades depending on different media characteristics. These
are expressed in the dimensionless power number N p (also called Newton number
relating the resistance force to inertial force) correcting the relationship between the
aforementioned variables and the real power demand.
The power number depends on the fluid flow pattern of a stirred tank bioreactor.
It can be described by impeller Reynolds number Reimp [−], indicating whether the
flow in the stirred tank is in the laminar, transitional, or turbulent regime. In general,
the Reynolds number (Osborne Reynolds, 1842–1912) is the ratio of inertial forces to
viscous forces within a fluid. So, the power characteristic (NP (Reimp )) is of practical
interest. For the given application it can be:
ρ liqu ⋅ Nimp ⋅ D2imp
Reimp = (5.2)
ηliqu
where ηliqu is the dynamic viscosity of the medium.
The correlation between the dimensionless numbers NP and Reimp depends on
the flow regime in the bioreactor. In a laminar regime, Reimp is less than 102 and the
power number is inversely dependent on Rei :
1
N p = k imp ⋅ (5.3)
Reimp
Using Equations (5.1) to (5.3), Equation (5.4) is obtained:
2
PR,imp = k imp ⋅ ηliqu ⋅ Nimp ⋅ D3imp (5.4)
98 | 5 Bioreactors – designing a home for the bioreaction

Still unknown influences are summarized in the proportionality constant k imp espe-
cially dependent on the type and number of impellers. Equation (5.4) indicates that
required power is independent of the density of the fluid for laminar flow but it is
dependent on the viscosity of the fluid. This case can happen for highly viscous fer-
mentation suspension e.g., during polysaccharide production.
In turbulent flow, Reimp > 104 and the power number is independent of Reimp :
3
PR,imp = NP,turb ⋅ ρ liqu ⋅ Nimp ⋅ D5imp (5.5)

Again, Np,turb depends on the geometry of the impeller.

Equation (5.5) indicates that for turbulent flow the required power is independent
of the viscosity of the fluid but it is dependent on the density of the fluid, in bioprocess
engineering usually in the range 1 [kg ⋅ m−3 ] < ρ liqu < 1.3 [kg ⋅ m−3 ]. In the transition
regime, which is in between laminar and turbulent flow, required power is dependent
on both density and viscosity of the fluid.
The presence of aeration induced bubbles results in a reduction in the power re-
quirement and transfer, since the gas bubbles decrease the density of the fluid and
affect the hydrodynamic behavior of the fluid around the impeller. The effect of aera-
tion, which is of course the usual case, upon power consumption is expressed as:
0.45
P2R ⋅ Nimp ⋅ D3imp
PR,gas = 0.72 ⋅ ( ) (5.6)
f 0.56

PR,gas is the gassed power, PR is the ungassed power, fGas is the volumetric gas flow
rate. For the above given correlations many variations are given in the literature con-
sidering many only qualitatively know influences like reactor geometry. In real appli-
cation the manufacturers of bioreactors give data sheets with real values. Some typical
values are given in Table 5.1.

Tab. 5.1: Typical values for power transfer and hydrodynamic parameters.

Cases V R [L] Nimp [min−1 ] Reimp [−] NP [−] PR [W ⋅ m−3 ] t c /t m [−]

1 0.1 1000 10 0.7

2 0.1 2000 5 3.0
3 10 300 2.5
4 10 500 5.0
5 2000 50
6 2000 200

1) Mini reactor, one stirrer, low speed 2) Mini reactor, one impeller, high speed; 3) lab reactor, one
stirrer, low speed 4) Lab reactor, three impellers, high speed; 5) medium size production; low stirrer
speed; blade; 6) medium size production; high stirrer speed. Find other values for different reactors
and impellers from supplier manuals.
5.2 Revisiting the stirred tank reactor – most important issues from chemical engineering | 99

Average energy dissipation rate per unit mass (ε [W ⋅ kg−1 ]) is defined as:
PR
ε= (5.7)
VR ⋅ ρ liqu
where VR is the liquid volume in the bioreactor and ρ liqu is the liquid density. It repre-
sents the power input per mass of medium.
Using Equation (5.5) and (5.11), Equation (5.12) is obtained:
3
N p ⋅ Nimp ⋅ D5i
ε= (5.8)
VR
After the mechanical energy is transferred from the stirrer to the medium it has to be
dissipated throughout the reactor. The local energy dissipation rate ϵ(x) over average
energy dissipation rate (ε) represents the dimensionless unit ε/ε. A knowledge of the
overall and the local distribution of energy dissipation is necessary for a successful
scale up, bubble break up or coalescence, estimation of shear stress, and many other
process applications. Now a closer look at the flow patterns will be visualized for un-
derstanding the basic action and impact of the agitators. This includes the flow profile,
the turbulence expansion and the shear rate, which can be understood as density of
the isolines of the flow profile.
A significantly high local energy dissipation occurs at the Rushton turbine espe-
cially at the blade tips. This causes stress for sensitive microorganisms. During scale
up the velocity of the stirrer tips increases due to higher circumferential speed, so
lower stirrer speed is necessary. This can be a limiting step. Therefore, the choice of
appropriate impeller is vastly important especially when shear sensitive microorgan-
isms are used (Figure 5.5).

Turbulence expansion [m2/s] and flow profile (arrow & flow lines)
2800 x10-2

2400
1.2
2000
1
1600
0.8
1200
0.6
800
0.4
400

0 0.2

-400
-2000 -1000 0 1000 2000

Fig. 5.5: CFD simulation of flow profile and turbulence expansion of a reactor with one Rushton tur-
bine. The unit m2 ⋅ s−1 becomes kinetic energy gradient per kg fluid.
100 | 5 Bioreactors – designing a home for the bioreaction

Turbulence expansion [m2/s] and flow profile (arrow & flow lines)
2800 x10-2
2400 1.6

2000 1.4
1.2
1600
1
1200
0.8
800
0.6
400 0.4
0 0.2
-400
-2000 -1000 0 1000 2000

Fig. 5.6: CFD result for an axially acting impeller.

Turbulence expansion [m2/s] and flow profile (arrow & flow lines)
2800 x10-2
5
2400
4.5
2000
4
1600 3.5
3
1200
2.5
800 2
400 1.5
1
0
0.5
-400
-2000 -1000 0 1000 2000

Fig. 5.7: CFD simulation of the impact of an Intermig agitator.

Lack of an axial component leads to a significantly low medium movement in the up-
per part of the reactor. This has to be compensated by several agitated layers.
Axial flow impellers create a top to bottom motion as Figure 5.6 demonstrates.
The local shear stress is much lower than for the six-blade impeller. Only one level
is necessary as fluid flow is directed first to the bottom and then bent to the top reach-
ing the whole reactor. The Intermig type impeller (Figure 5.7), compared with the Rush-
ton turbine, transfers more uniform energy to the fluid and accordingly demands less
power consumption in order to obtain the same degree of mixing and mass transfer
coefficient accompanied by lower shear stress.
5.2 Revisiting the stirred tank reactor – most important issues from chemical engineering | 101

Efficient mixing is necessary to care for an ideal homogenization of all com-

pounds. This is to prevent sedimentation of the cells, transportation of gas from the
medium to the bubbles and vice versa. Not unimportantly, heat has to be brought to
the reactor wall by convection. Fresh medium and titration agents are usually dosed
locally and have to be suspended, otherwise local gradients would occur changing
cell physiology.
Mixing efficiency of stirred tank reactors are evaluated by mixing time. Short mix-
ing times are necessary for a proper bioreactor design. Mixing time, tM , denotes the
time required to achieve a certain degree of homogeneity, which is a measure of the
quality of a mixture. The mixing time measurements are performed by injecting a
tracer into the stirred vessel. Tracers mainly include acidic or alkaline solutions, salt
solutions, heated solutions, and colored solutions. Tracer concentration is measured
at fixed points by various methods such as sensor method, decolorization method,
and optical and colorimetric methods.
The mixing time depends on liquid properties, the stirrer type and size, the spe-
cific power input, and bioreactor geometry. At the required degree of homogeneity (M),
tracer concentration reaches within ±5% range of final concentration (cf ) after fluctu-
ation. Circulation time (tC ) is defined as the time for the fluid in the stirred tank vessel
to circulate one loop. The degree of homogeneity M is then defined as:

∆c ct − ci
M = 1− = (5.9)
∆c0 cf − c i

where cf is the final/equilibrium tracer concentration, c t is the tracer concentration at

time point t, and c i is the initial tracer concentration. ∆c0 is the concentration differ-
ence at the beginning of mixing and ∆c is the concentration difference at respective
time t. The temporal course of the concentration fluctuations can be approximated by
an exponential approach as in Equation (5.10):
∆c tM
= k M ⋅ exp (− ) (5.10)
∆c0 tC
where k M is a constant, tM is the mixing time and tC is the circulation time. Mixing
time can also be expressed in dimensionless form in a turbulent flow regime:

N i ⋅ tM = NM = const. (5.11)

where N i is the rotational speed of the stirrer and NM is the mixing number. Mixing
number reveals the number of stirrer rotations required for a specific degree of ho-
mogenization. The course between mixing number and impeller Reynolds number is
called mixing time characteristics and can be created for various stirrer types. In a lam-
inar flow regime, NM decreases with increasing impeller Reynolds number whereas in
a turbulent flow regime, when Re>104 , mixing number NM is constant, having no re-
lation with impeller Reynolds number.
102 | 5 Bioreactors – designing a home for the bioreaction

The last point to discuss is heat transfer. Bioreactions are generally exothermic.
We remember that in aerobic bioreactions half of the substrate is completely oxidized
in respiration. So a heat balance can be set up:

∆cGluc ⋅ HC,Gluc = Y X,Gluc ∆c X ⋅ HC,X + Qheat (5.12)

Therefore, for the volumetric heat flux qheat :

qheat = rGluc ⋅ c X ⋅ (HC,Gluc − y X,Gluc ⋅ HC,X ) (5.13)

For an example we set rGluc = 1 g/(g ⋅ h), c X = 50 g/L, y X,S = 0.5 g/g, HC,Gluc =
HC,X = 16 MJ/kg we obtain qheat = qheat = 400 kJ/(L ⋅ h). That means that the medium
would heat up by 100 °C (cP,H2O = 4 kJ/(kg ⋅ K). Indeed, high temperatures can oc-
cur in haystacks and even self-ignition, induced by microbial activity. Silos filled with
organic powders have also been reported to burn in cases where humidity has led to
microbial contamination. Discharge of heat in bioreactors goes along convective trans-
port to the reactor wall and then through it by thermal conduction into the cooling
liquid. With the help of Fourier’s law this process can be described:
Tmedium − Tcool
qheat = k heat ⋅ Awall ⋅ (5.14)
Dwall
where Awall is inner surface of the reactor contributing to heat transfer, and k heat is the
material’s heat conductivity [W ⋅ m−1 ⋅ K−1 ]. In practice, the reactor wall is not a plane
but has different inside and outside surface area. Together with surface roughness this
means that dwall cannot be exactly determined. A laminar medium film complicates
the problem further by making heat transfer dependent on medium turbulence. The
final heat transfer from the medium into the wall material is described by the overall
heat transfer coefficient. The material is not known by the operator and heat radia-
tion may contribute. So a practical value is provided by the manufacturer, the thermal
transmittance U = k heat /dwall [W ⋅ m−2 ⋅ K−1 ]. This lumping of influences and replac-
ing exact values by lumped experimentally obtained values is a common approach in
chemical engineering.
Cooling jackets are preferred in pilot and medium size production reactors as they
do not expose additional surface area (biofilm formation), do not reduce active work-
ing volume and facilitate cleaning. In large reactors heat exchangers are sometimes
not avoidable. This also holds for small glass lab reactors, where a ‘cooling finger’ can
be positioned. Glass exhibits only 1% of the heat conductivity. This has also impact on
lab work. Cooling down a shaking flask in an ice bath may happen only with 1 °C/s,
which is too slow for some purposes.
Finally, we come back to mechanical power transfer, where another specific issue
of bioreactors has to be envisaged belonging to the topic of hygienic design. To pre-
vent leakage between shaft and vessel a mechanical seal is built in as a shaft bearing;
details are made clear in Figure 5.8. This type of seal restrains the medium via two
5.2 Revisiting the stirred tank reactor – most important issues from chemical engineering | 103

Inside Spring
Clamp ring
Reactor
boom
O-Ring, dynamic

Outside Rotary ring

Staonary ring

Seal liquid O-Ring

(a) (b)

Fig. 5.8: CAD drawing of a double-acting seal, (a) overview, (b) details.

rings, from which two surfaces are in direct contact with each other. One ring turns
with the shaft and is called the rotary ring; the stationary ring is fixed at the reactor
vessel. A spring element presses the two rings against each other thus forming the el-
ements of a face seal. Medium forms a lubricating film between the surfaces when the
shaft rotates. The pressure in this film balances the seating force. Next to this primary
seal an additional secondary seal formed by O rings (toric joints) has to be foreseen.
Only one mechanical seal would allow microorganisms to escape to the outside or to
enter the reactor via the lubricating film. The double mechanical seal with one at the
product side and one at the outside allows for application of sealing liquid or steam
for deactivation of cells approaching the room between the two seals.
To avoid all these problems, magnetic coupling between the outer engine and an
inner shaft has also been developed. For small scale this principle is already known
in the form of a magnetic stirrer using a magnetic stir bar. Applications can also be
found for smaller bioreactors. For large dimensions magnetic coupling has not been
widely established.
The STR can be universally applied and is in the lower and middle scale among the
most common fermentation vessel. Some disadvantages have to be mentioned as well.
Firstly, energy dissipation rate is very high at the tips of the stirrer blades. The second
point is that axial mixing is worse than radial mixing. These points are addressed by
the draft tube reactor (Figure 5.4), where the fluid is forced in the axial direction by a
propeller, then bent by the dished ground plate (torispherical dished end), and finally
recirculated to the top. The occurring vortex can also suck foam back into the fluid.
This reactor is especially of advantage for sensitive cells at lower oxygen transfer rates.
104 | 5 Bioreactors – designing a home for the bioreaction

5.3 Pneumatically driven bioreactors – a soft way of energy

transfer

But who said that stirrers are the only means to make something move? We can think
more generally of ways to transfer mechanical energy into the medium and care for
adequate flow patterns. Forces on the technical scale can be easily applied either by
pneumatic or hydrodynamic pressure. Flow control is gained by static installations
adjusted to the energy transfer and the specific needs (Table 5.2).

Tab. 5.2: Possible means of mechanical energy supply and appropriate flow pattern control.

Energy transfer/ Stirrer Gas bubbles Pump

Flow control (mechanical) (pneumatic) (hydraulic)
Baffles Stirred tank reactor – –
Draft tube Draft tube reactor Air lift reactor Jet reactor
None Cup of coffee Bubble column Jet reactor

Bubble column bioreactors are pneumatic reactors in which aeration and agitation
are enabled by gas bubbling without using any mechanical stirring device. Therefore,
bubble column bioreactors in general have no internal structures except a gas sparger,
which is usually shaped as a perforated plate. In the simplest case bubble columns are
tubes standing upright. A typical aspect ratio HR /DR = 10 or even higher allows for
a long residence time of the bubbles. While the columns can be up to 100 m high,
the working volume can be 300 m3 or even up to 3000 m3 . Occasionally when tall
columns are preferred, perforated horizontal plates are used to break up the coalesced
bubbles (Figure 5.9).
Bubble column performance is influenced by the hydrodynamic behavior of the
bubbles. Sparger design, medium viscosity, column diameter, and especially gas flow
rate have an important influence on the development of different flow regimes. As the
introduction of gas at the bottom is the only energy input specific attention is required.
In contrast to the STR gas flow FG is not normalized to the volume but is based on the
cross-sectional area AR . This approach delivers a superficial gas velocity UG [m ⋅ s−1 ]:
FG
UG = (5.15)
AR
The idea behind it is that rising bubbles evoke similar effects along the longitudinal
axis of the reactor with respect to energy and mass transfer. At low superficial gas
velocities (≤ 0.05 m ⋅ s−1 ) bubbles have uniform shape and rise through the column
without interacting with other bubbles and thus a homogeneous flow pattern occurs.
With the increase of gas flow rate, e.g., up to 1 m ⋅ s−1 , a heterogeneous flow regime
occurs due to the unstable and turbulent motion of the bubbles. This leads in particu-
lar to bubble dispersion and coalescence. Here large bubbles emerge. This can even go
 5.3 Pneumatically driven bioreactors – a soft way of energy transfer | 105

Air outlet Air outlet

Headspace
(Foam)

Perforated
intermediate Dra tube
plates

Sprinkler

Air inlet Air inlet

(a) Bubble column reactor (b) Airli reactor

Fig. 5.9: Scheme of the two basic designs of pneumatic columns reactors, (a) bubble column, (b) air-
lift reactor. Bubble coalescence occurs on the way up, perforated plates lead to dispersion again.
Bubbles size will also increase due to decreasing hydrostatic pressure.

so far that the bubble diameter reaches the diameter of the reactor. The rising bubble
leaves no space for the backflow of the fluid. This phenomenon, known as slug flow, is
sometimes intended in small tube systems, as the bubbles press liquid layers through
the tubes. In larger reactors it is dangerous. Cases are reported where slug flow even
led to destruction of the reactor. In the homogeneous flow regime gas holdup ϵG is
proportional to UG , which is physically intuitive assuming constant and same rising
velocity of all bubbles. In the heterogeneous regime ϵG shows a disproportionately
low increase with UG .
Since no mechanical stirring device is applied, bubble column bioreactors require
less energy or to say it the other way round do not allow for high energy and mass
transfer compared to the STR, comprising an agitation system. The volumetric power
input exerted to the fluid in the bubble column depends on the superficial gas velocity
and can be determined by Equation (5.16):
PR
= ρL ⋅ g ⋅ UG (5.16)
VR

where PR [W] is the power supplied for aeration, VR [m3 ] is the liquid (working) vol-
ume, ρ L [kg ⋅ m−3 ] is the liquid density, g [m ⋅ s2 ] is the gravitational acceleration con-
stant, and UG [m ⋅ s−1 ] is the superficial gas velocity. The rationale behind this is that
the bubbles have to overcome the hydrostatic pressure at the bottom of the reactor.
The locally distributed values are shown in Figure 5.10 as simulation results.
106 | 5 Bioreactors – designing a home for the bioreaction

Speed profile of the liquid phase [m/s] Shear rate [log (1/s)] Volume fracon of the gas phase [%]
80
x10-2
2.5 5
0.3 4.5
60 60 2
60 4
0.25 1.5
3.5
40 40
0.2 1 40 3

2.5
0.5
0.15
20 20 2
20
0
0.1 1.5

-0.5 1
0 0 0
0.05
-1 0.5
-5 0
z 5
-20 -20 y x -10 -5 0
0 20 0 20

Fig. 5.10: CFD simulations of some hydrodynamic parameters of the bubble column.

Low capital cost, gentle environment for sensitive microorganisms, easy cleaning of
the vessel, and lack of mechanical stirring device are the main advantages of bub-
ble column bioreactors, while low energy transfer and accordingly mass transfer are
limitations. Applications in industry that fit this profile include cultivation of filamen-
tous fungi for citric acid and penicillin production. Later we will see that low growth
rates, e.g., 0.1 h−1 for Aspergillus, is accompanied by a low oxygen consumption rate.
Beside the advantages, disadvantages such as foaming and the formation of coales-
cence induced gas bubbles decreases the bioreactor performance. The high pressure
at the bottom may not affect the microorganisms directly but influence the CO2/HCO3
balance and therefore pH value. Low axial mixing time is another problem, which is
tackled by a reactor design described in the next paragraph.
Airlift bioreactors are likewise pneumatic tower shaped reactors, in which agita-
tion is enabled by gas sparging. The design of the airlift bioreactors includes a draft
tube, which is situated in the center of the reactor. The draft tube creates two distinct
zones in the reactor where only one of these zones are sparged using a gas. The den-
sity difference between the aerated zone, the riser, and the not aerated zone, the down-
comer, results in a pressure difference between the two columns and subsequently to a
liquid circulation allowing an adequate mixing. In the riser gas-liquid upflow emerges
while in the downcomer a fluid downflow exists (Figure 5.9 (b)). This principle was al-
ready invented 230 years ago for pumping suspensions.
In general, airlift bioreactors have two basic configurations. These are internal
loop airlift bioreactors and external or outer loop airlift bioreactors. In the internal
loop airlift bioreactors, the bubble column is separated by an internal baffle so that a
riser and a downcomer occur, whereas in the external or outer airlift bioreactors, riser
and downcomer are separate tubes connected by short horizontal sections at the top
 5.3 Pneumatically driven bioreactors – a soft way of energy transfer | 107

and the bottom. External loop airlifts have a better mixing efficiency and a faster liquid
circulation since the density difference between fluids in the riser and downcomer is
greater as a result of extended distance.
Due to the lack of a mechanical agitator system, construction cost and energy
consumption accordingly reduce, less maintenance is needed, and an easier steril-
ization is possible compared with the bioreactors that contain agitator shafts. More-
over, despite the advantages of airlift bioreactors, the necessity of relatively higher
pressures, greater air throughput and the inefficient foam breaking should be consid-
ered depending on the process requirements. Owing to the controlled liquid flow and
equalized shear forces, airlift bioreactors have high efficiencies opening new possi-
bilities in aerobic bioprocessing technology. Therefore, airlift bioreactors are used in
many applications such as the growth of shear sensitive cells, the production of single
cell proteins and immobilized enzyme reactions.
The volumetric power input exerted to the fluid is defined by the degree of turbu-
lence and can be determined by using Equation (5.17):
PR ρL ⋅ g ⋅ UG
= (5.17)
VR 1 + Ad Ar

where A d [m2 ]
is the cross-sectional area of the downcomer and A r [m2 ] is the cross-
sectional area of the riser.
The third means of energy transfer, namely hydrodynamic pumping, is used
mainly in the wastewater field to increase residence time in the reactor while main-
taining good mixing. Apart from that the use in biotechnological processes is limited.
The basic schemes can be taken from Figure 5.11.

Gas outlet Gas outlet

Free jet reactor

Dra tube

(a) Gas inlet (b) Gas inlet

Fig. 5.11: Jet reactors (a) with (free jet) and (b) without draft tube.
108 | 5 Bioreactors – designing a home for the bioreaction

5.4 Other types of bioreactors – translating demands into design

Standardization is a general demand in bioreactor design to keep things simple and

cheap. Nevertheless, specific designs for specific applications can exhibit significant
advantages to justify tailored designs. Gradual improvements with respect to aspect
ratio, specific stirrers (e.g., to cope with high viscosity) or operational parameters
mark one direction. Another thing is to introduce additional degrees of freedom. Until
now medium, cells, and dissolved gases have been treated as ideally mixed. Bubbles
are an exception as they experience additional buoyancy forces. This is not optimal
as the residence time cannot be exactly adjusted to fulfill the task of oxygen supply
or mixing simultaneously. Decoupling of solid phase (cells), medium (inclusive dis-
solved gases), and gas phase with respect to transport means and patterns is wor-
thy of consideration. Gas could also be applied by tubes or membranes, and medium
and/or cells may be really solid giving room for application of gravitational or struc-
tural forces. These could be fibers or granules as substrate and immobilized cell pellets
or biofilms for the biomass. Table 5.3 is an attempt to structure the different possibili-
ties.

Tab. 5.3: Possible ways shaping transport processes of gases, medium and biomass.

Gas or substrate supply / Surface Bubbles Membrane Liquid film

Substrate, biomass
Dissolved Shaking flask Submers Membrane reactor
(regular case)
Solid Emers reactor Biomass pellets Biofilm Trickle bed

Now a few reactor types will be picked out and sketched.

Before agitated bioreactors came into use biomass e.g., filamentous fungi propa-
gated by swimming on the surface of the medium. In this way the mycelium could take
substrate from below while oxygen uptake was enabled by the air above the medium
surface. Citric acid as a product was excreted into the medium. This surface cultiva-
tion is called emers cultivation in contrast to the nowadays usual submers cultivation.
Production is possible in simple flat stackable vats without any auxiliary equipment.
Sometimes this approach is still employed for spore germination. Some plants also
need direct contact with air for propagation. The degree of process intensification is
of course very low.
Biomass can also form a more solid phase as pellets (filamentous fungi), or cell
flocs (wastewater). These aggregates are large enough to be treated as a separate phase
independent from medium hydrodynamics. This cannot only be utilized for cell har-
vesting but also for retaining the cells in the bioreactor by gravity or structural means
like sieves (compare Chapter 9, biomass recycle). The most widespread application
 5.4 Other types of bioreactors – translating demands into design | 109

is the upflow anaerobic sludge blanket reactor (UASB). Wastewater enters the reactor
from the bottom via a jet or nozzle and flows upward. The upward fluid velocity in the
reactor is kept low enough to allow the biomass pellets to sediment against the flow
with the same velocity. Gas bubbles from anaerobic metabolism (biogas) care for ad-
ditional turbulence. The gas is separated at the top or at several layers by deflectors
(baffles) and a gas collector (caps, domes). In the case of low upstream velocities axial
gradients of substrate concentration may emerge. This effect may even be intention-
ally supported. The effect is that easily metabolized medium compounds are used up
at the bottom while the microorganisms at the top are forced to take up the more re-
sistant compounds. Such a gradient also occurs inside the pellets leading to different
bacterial consortia long the gradient. So the message is that the continuous stirred
tank reactor (CSTR) comprising ideal mixing is not always the best choice. A perfo-
rated plate can keep them back in case higher medium flows are possible, reducing
axial gradients. In this case partial medium feedback can be envisaged via the nozzle,
an application of the jet reactor principle.
Autonomously arising flocs can be regarded as biofilms, where the cells attach
to each other either passively or actively. Biofilms in the narrow sense grow on the
surface of solid substrates, here meaning an inert matrix. In such a biofilm the cell
density and the activity can be much higher than in suspension culture, especially
in anaerobic processes. The substrate can be formed as small (cm) parts called carri-
ers e.g., from plastic with high surface area. A popular shape is a spoked wheel. Such
overgrown carrier can sediment much faster allowing higher flow rates. The moving
bed biofilm reactor (MBBR) improves reliability, simplifies operation, and requires less
space than traditional wastewater treatment systems. In addition, slowly growing and
sensitive cells like mammalian cell lines can be cultivated using this principle. The
cells do not form a biofilm as such but settle in pores of a porous carrier. In a trickle
bed reactor the carriers do not flow freely but form a packed bed. The medium does
not flow through the bed upwards but is trickled from above. The gas phase is now
the continuous phase allowing for good and controllable gas transfer via diffusion
through the extensive water film on the carriers. This principle can therefore also be
applied to aerobic processes. Note that a composter is also a bioreactor with biofilms
growing on the solid substrate and kept wet in a trickle bed. To make use of cellulosic
wastes, trickling hay bales with enzyme solutions have been tried out on a large scale.
A great future will be awarded to biofilm reactors. Current research is in the direction
of completely controlled biofilms. One means is the application of substrates either
by a fluid medium or by diffusion through a membrane that is the substrate. Sponta-
neous biofilm formation is replaced by molecular linkers attached on one side to the
substrate, and on the other to a specific epitope of the cells. In this way only strains
of interest are part of the biofilm, while contaminants are washed out. Zero growth
conditions are maintained without washout and carbon is mainly assigned to an ex-
tracellular product.
110 | 5 Bioreactors – designing a home for the bioreaction

Besides mass transport issues, other very important aspects of reactor design in-
clude the interconnected aspects of scale, economics, product and microbial physiol-
ogy. In general, larger reactors are cheaper than small ones calculated on a volume
basis. Only considering the expenses for stainless steel the reactor surface ratio leads
2/3
to cost ∞VR . Measurement equipment is necessary only once. Unfortunately, stirred
tank reactors have their limit of scale in the range of a few thousands m3 . For high
value products like pharmaceutical proteins, quite expensive reactors are needed to
fulfill safety and quality issues. This generates high volume specific costs, but these
are justified by the high market price, which is calculated also based on downstream
operations and last but not least on payback of development costs like clinical studies.
For cheaper bulk products like citric acids or penicillin very high reactor volumes are
necessary. This makes employment of bubble columns sensible, even if compromises
in oxygen transfer rate are unavoidable. This is made easier insofar as fungal culti-
vations need lower oxygen transfer due to lower growth rates but benefit from lower
shear rates. The bottom of the value chain is represented by wastewater treatment
and anaerobic digestion. Here nobody speaks about stainless steel or axenic opera-
tion. Reactors are made from metal sheets or concrete. Peripheral devices are kept to
a minimum.
Now we should spend some time considering how to design, calculate, and decide
dimensions for such large plants. Scale up follows the usual rules: first describe lim-
iting parameters and then calculate their development during scale. In bioprocesses
this is not that easy. While calculation of things like energy transfer and flow patterns
may be possible using common correlations or CFD simulations, cell reactions in re-
sponse to the environment in particular are not known a priori. The situation is ab-
stracted in Figure 5.12.
On the outer level the production goals are defined, e.g., high productivity or high
product concentration. Direct access to the actual transport rates of substrates, gases
or energy flows is possible by valves and pumps. On the medium level design flow
patterns can be specified by internal installations like baffles and mixers. Calcula-
tions using the transport level parameters is done by dynamic balance and reaction
equations. However, some parameters like turbulence or diffusion can only be roughly

• Transport
Reactor • Control

• Flow paern
Suspension • Gradients
• Phase separaon
• Physiology
Cell • Metabolism
• Genecs Fig. 5.12: Situation bioreactor calculation
on three different layers.
 5.4 Other types of bioreactors – translating demands into design | 111

estimated. On the cell level we have to face the dominant inaccuracy with respect to
kinetic parameters. Kinetics is the interface between the cell and the medium level.
Consequently, the design process has to go the other way around: first determine these
parameters, then calculate and define optimum conditions in the biosuspension. Here
limitations like mass transport by diffusion (e.g., bubbles) can be rate limiting. Finally
the necessary external transport parameters have to be fixed. No additional limitations
should occur on this level but be avoided by design. So, we need small bioreactors to
measure layout parameters for big ones. This approach is called ‘scale down’. In the
following examples we will follow the read thread from very small to very big.
The highest specific costs result if scientific data are the product itself. Strain char-
acterization or medium design are typical examples of scientific goals to be achieved
in bioreactors. Scanning the possible parameter space can lead to a combinatorial ex-
plosion. To cope with this situation very small reactors on the µL scale can be designed
allowing for a high degree of parallelization. In small scales diffusion contributes pos-
itively to homogenization. An example is shown in Figure 5.13.
The cells under investigation are kept in a small chamber and are separated from
the upper part by a membrane. Fresh medium is supplied in this upper part, while
compounds for supplying the cells can diffuse through the membrane. To fulfill the
desired goal of data acquisition, sensors-on-a-chip and microsensors are employed.
Microscopy as an information source is not out of fashion but foreseen here by optical
windows. Note that neither bubbling nor mechanical mixing is necessary. To increase
experimental throughput microtiter plates (µL scale) are employed. Fast screening of
strain libraries is the preferred task to be accomplished. In the best case, gas exchange
is possible but no medium exchange. Growth can be observed by an optical reader
giving data for biomass, pH, pO2 , or fluorescence. Nevertheless, chemical analysis is
possible only at the end of a batch run. So the results cannot really be extrapolated
to anticipated process conditions. The last step towards small (pL) scale is the adap-
tation of microfluidic devices making single cell analysis possible. This opens a new
dimension in process development. In bioprocesses we observe only medium growth

Biochip
Opcal window Semipermeable
Microbioreactor´s case cover membrane

Microsensors Microsensors

Inflow Oulow

Microsensors
Biological Biochip Opcal window
material

Fig. 5.13: Microreactor for investigation of cell growth (© CytoBioTech).

112 | 5 Bioreactors – designing a home for the bioreaction

characteristics of all cells. In reality there is a heterogeneity in physiological param-

eters depending on the point in the cell cycle or other partly unknown reasons. This
holds even for monoclonal cultures. Single cell analysis offers an elegant way to follow
e.g., physiological changes during the cell cycle of yeast. Synchronized cultivations
are an expensive alternative delivering only qualitative data.
Back at the mL scale we find the shaking flasks as stepping stone in process de-
velopment. The specific power input (P/V [kW ⋅ m−3 ]) depends mainly on shaking fre-
quency. Typical values are around 1 kW ⋅ m−3 but can be increased up to 3 kW ⋅ m−3
for shaking frequency of e.g., 300 min−1 . This is actually on the order of magnitude of
stirred tank reactors. Small filling volumes and baffles can increase the number fur-
ther. Large flasks (e.g., 2000 mL) can reach even higher numbers up to 6 kW ⋅ m−3 .
With respect to oxygen transfer the sealing cap could be a bottleneck, but headspace
aeration with small tubes is also an option. Measurements of pH and pO2 are possi-
ble with internal optical sensor pads. Color changes are read out via light beams. The
shaking flasks is really a suitable tool for multiple parallel experiments towards pro-
cess development.
A better option for observation of cultivation are parallel minireactors (about
200 mL scale) allowing for control of the most important parameters and even for
fed-batch processes. One commercially available reactor system is shown in Fig-
ure 5.14. This is a compromise between parallelization and process similarity. There
are sets on the market with four or more (‘six pack’) such minireactors sharing the
same peripheral devices. Several batches with different parameters are possible in

Fig. 5.14: Dasgip multiparallel mini reactor installation with process and control periphery included
(© Eppendorf). Visible are the magnetic stirrer through the headspace, off-gas cooling and mem-
brane vent filters.
 5.4 Other types of bioreactors – translating demands into design | 113

Fig. 5.15: Wave rector for animal cell

culture with several disposable plastic
bags (© GE healthcare).

direct comparison. In addition, sterilization a high degree of automation and data

acquisition is possible.
Not only for James Bond, but also for sensible cells and high value products
‘shaken not stirreds is mandatory. Animal cells do not have a rigid cell wall and
cannot stand strong agitation. Here reactors have been developed that realize non-
invasive mixing by smoothly rocking the whole reactor. One example is shown in
Figure 5.15. The plastic bag (‘cell bag’) contains the cell suspension, while the sup-
port (rocker) carries out slow rocking (rolling or pitching) movements. This induces
wave-like movements of the fluid, which gave rise to the generic name as propagated
by a company of ‘wave bioreactor’. Oxygen transfer happens via the headspace. The
working volume from less than 1 L up to several 100 L is even good for production of
small charges of high value products like antibodies.
In the last picture the cultivation is actually a plastic bag. This is meant to be dis-
posed of after use. Sterilization is done by radiation during the manufacturing process
making thermal sterilization superfluous, a big advantage. Plastic bags are of course
cheaper than stainless steel reactors. Changing the bag after each cultivation allows
for higher flexibility producing small charges for the preclinical and clinical market.
Current discussion is around whether this really pays back. The main parameters dis-
cussed are duration of a batch and the added value of the product. The application
of single use bioreactors (SUBs) in biopharmaceutical research and production has
increased enormously in recent years. They are employed for production of biophar-
maceuticals like vaccines and biosimilars in all primary process steps, in particular for
small and medium volume scales (Figure 5.16). SUBs are primarily used in processes
114 | 5 Bioreactors – designing a home for the bioreaction

Fig. 5.16: Complete ready to use unit using disposable bags, here in the left steel vessel. Further
downstream processing like harvesting and extraction units can be designed as single use equip-
ment in the same rack (© Sartorius).

in which protein based biotherapeutics from mammalian cell cultures are the target
product. Additionally, they can be used for cultivation of plant cell cultures, microor-
ganisms, and microalgae, as well as for special products in the food and cosmetics
sector. A large variety of single use bioreactors and single use mixing systems with
volumes of up to 2,000 L is presently on the market. These systems differ in terms of
mixing, type of power input, and gassing strategy. An outer support container is en-
gineered and fabricated to fully support each SUB fermentation bag and allow easy
access for operation. It contains the mixing drive, possibly a silicone electric resistive
heating blanket or water jacket, and optional controllers for mixing.
Having said already a few things about standard reactors we now make a jump
to really big vessels but employed for low value products. The ‘biotower reactor’ for
aerobic wastewater treatment carries this direction to extremes. Figure 5.17 shows a
picture and the internal structure of this reactor. The reactor should not be mixed up
with an anaerobic digester. It is basically a bubble column, but with immediately vis-
ible differences. First of all the reactor is not tall as expected for bubble columns. The
air does not leave the sparger upwards but downwards against the wastewater stream
being of course the medium. This measure leads to a first gas dispersion. A shield leads
the bubbles to the outer regions of the reactor where they can freely rise. The bubble
column effect appears now in the opposite direction compared to the usual geometry
of bubble columns. Such a ‘tower biology’ needs more pumping energy and is more
 5.4 Other types of bioreactors – translating demands into design | 115

Exhaust air

Clear
water

Secondary Surplus sludge

clarificaon
Return sludge
Acvated
sludge Air

Liquid phase
(a) (b)

Fig. 5.17: Appearance (a) and sketch of the internal structure (b) of the biotower reactor (© in-
fraserve).

Fig. 5.18: Virtual reconstruction of an ancient fermentation jar, the small ‘chimneys’ are probably
designed as outlet for CO2 keeping the foam back in the vessel.

expensive than an activated sludge tank while the main advantage is space saving. In
fact, the reactor integrates a whole wastewater plant, especially as it comprises also
sludge separation. This is done in the ring shaped structure at the top. Sludge can sedi-
ment (flocculation occurs naturally) and can be removed or recycled while the cleared
water is discharged. The biotower reactor is therefore an example for integration on
the process level.
Back to our roots: What did our ancestors do to produce ethanol without stainless
steel or disposable plastic bags? Beer and wine were usually produced in amphorae
or inconspicuous ceramic pots. Sometimes archeologists cannot even decide whether
it is a jar or an urn without investigating the content by modern analytics. Maybe,
bioprocess engineers could sometimes contribute to this integration of sciences. An
especially nice ancient fermentation vessel is shown in Figure 5.18.
116 | 5 Bioreactors – designing a home for the bioreaction

5.5 Gas transfer – supplying microorganisms with gaseous

compounds

Beside dissolved organic substrates and minerals, gaseous compounds also have to
be supplied to the microorganisms as educts. With respect to biochemical engineer-
ing gas supply requires special consideration and technical implementation. Low sol-
ubility requires constant feeding. Gases are also endpoints of biochemical reaction
chains and have to be removed out of the reactor. The transport of dissolved gases
into the medium is a widespread reason for limitations in process productivity. The
final strived product can also be a gas. This makes in situ product removal easy and
can potentially reduce downstream costs. All these facts are reason enough to have a
closer look at the biological, physical, and technical aspects of gas turnover in biore-
actors.
Oxygen (O2 ) is needed by most aerobic microorganisms for respiration and almost
only for this purpose. The cells take it up from the medium similarly to most gaseous
compounds by diffusion. Usually it is supplied to the reactor as part of normal air,
but also pure O2 can be employed in the case of a high oxygen demand or in order to
reduce bubble volume. As pure oxygen is much more expensive than air, the decision
to apply it depends of course also on the product value.
Carbon dioxide (CO2 ) is a product of respiration and has to be removed from the
reactor. To some degree carbon dioxide is fixed in anaplerotic sequences, e.g., for clos-
ing the tricarbonic acid cycle. There are some hints that strong stripping leads to ex-
tension of the lag phase. Phototrophic organisms need CO2 for photosynthesis and in
this case oxygen has to be removed. Further on, CO2 has side effects on the medium
as it leads to acidification.
Nitrogen as N2 is usually not involved in technical bioprocesses, so together with
the noble gases it is referred to as inert gas. Being part of normal air it is fed into the
reactor and contributes to gas exchange and mixing. Some anaerobic bacteria and
some phototrophs can reduce nitrogen to ammonia. As about 1% of the worldwide
energy is used to convert atmospheric N2 to NH3 in the Haber–Bosch process it is a
worthwhile goal to find a feasible bioprocess to perform this nitrogen reduction. Inert
gases are also used to strip traces of oxygen from the medium for cultivation of strictly
anaerobic organisms.
Methane (CH4 ) is produced in biogas plants along with CO2 by methanogenic mi-
croorganisms. Energy for mixing and product separation is provided by the bubbles
inclusively, an absolute ‘must’ in bioenergy production.
Hydrogen (H2 ) is also involved in biogas formation as an intermediate compound.
How to stop the process at that point producing hydrogen instead of methane is an
open question in ongoing research. Producing hydrogen by water splitting employing
microalgae is another option for the future.
 5.5 Gas transfer – supplying microorganisms with gaseous compounds | 117

Some volatile organic compounds could also possibly act as substrates or are
products like volatile flavors. Ethanol and acetaldehyde are products in alcoholic fer-
mentation and can be subject to desired or undesired stripping out. The bulk chemical
isoprene is discussed as a gaseous product from bioprocesses avoiding energy con-
sumption for cell separation and disruption.

5.5.1 Transport processes – the journey of gases through the reactor

Gas transfer to the reactor and into the medium can be understood as a sequence of
different transport steps as shown in Figure 5.19. Gas enters the reactor with the flow
FGas,in [L ⋅ h−1 ], where values are given for normal conditions. Related to the working
volume of the reactor volume, the volumetric gas flow fGas,in = FGas,in /VR [L ⋅ L−1 ⋅ h−1 ]
is obtained. In technical environments this is often given as gas volume per reactor
volume per minute [vvm]. The content of a particular gas G in the gas mixture is given
as molar fraction xG [mol/mol]. As an example, for aerobic cultivation, air enters the
reactor via a gas pipe ending in a sparger. This is in simple cases a tubular ring with
pores to form and release gas bubbles. The bubbles rise up driven by their buoyancy.
Due to the turbulences evoked by the stirrer the bubbles are dispersed in the reactor.
Oxygen diffuses through the bubble/medium interface and is more or less homoge-
neously distributed by convective transport. The volumetric rate with which oxygen
enters the fluid phase is called the oxygen transfer rate (OTR [g ⋅ L−1 ⋅ h−1 ]). The cells
are in direct contact with the medium and take up the oxygen by diffusion through
the cell membrane. The amount of oxygen consumption by biomass is calculated as
oxygen uptake rate OUR [g ⋅ L−1 ⋅ h−1 ]. Carbon dioxide leaves the reactor in the op-
posite direction out of the cells also by diffusion with the carbon dioxide production
rate (CPR), through the medium and into the bubbles with the carbon dioxide transfer
rate (CTR). Finally the bubbles burst at the surface of the medium and release the gas
into the headspace, from where it is set free into the environment with the flow rate
FGas,out .
The oxygen balance in the medium finally reads:

dc O2
= OTR − OUR (5.18)
dt

In the following paragraphs a closer look at the different transport steps is outlined.
To understand the route of oxygen out of the bubble into the fluid we have to revisit
two basic physical laws. The first one is Henry’s law (5.19), which gives a linear relation
between the partial pressure of a gas in the gas phase and the concentration in the
fluid phase. A prerequisite is that both phases are in direct contact and equilibrium
conditions are adjusted (Figure 5.20).
118 | 5 Bioreactors – designing a home for the bioreaction

Offgas
FGas,out Diffusion Convecon Diffusion

OUR OTR
Gas Phase
Solid Phase
Bubbles
Cells
VG xGas,g pG,g
CTR CPR
Aeraon Fluid Phase VR,l cG,l pG,l
FGas,in Medium

Fig. 5.19: Exchange of oxygen and carbon dioxide between bubbles, medium, and cells. The oxygen
transfer rate (OTR) from the gas phase to the liquid phase is driven by the concentration difference,
while the oxygen uptake rate (OUR) is controlled by the cells.

pG,Gas pG,Mem pG,Liq

Increasing gas
uptake
cG,Gas c*G,Liq
cG,Mem
cG,Liq
Gas phase Liquid phase
ideally mixed Membrane ideally mixed

Fig. 5.20: Course of partial pressure and concentration in the gas phase and liquid phase separated
by a membrane. Partial pressure for equilibrium conditions is the same in the gas and the liquid
phase but drops with increasing uptake by the cells. Concentration in the membrane can be different
due to a different solubility.

The equilibrium partial pressure of a gas G in the gas phase is given as:

pG,Gas = xG,Gas ⋅ PGas (5.19)

Which leads to the saturation concentration c∗G,Fluid in the liquid.

c∗G,Liquid = HG,Liquid ⋅ pG,Gas (5.20)

In this linear relationship the Henry coefficient HG,Fluid is a proportionality factor.

Some important Henry coefficients are listed in Table 5.4. Note that Henry coefficients
are dependent on temperature or salt concentrations in the case of a medium, so they
have to be looked up in chemical engineering handbooks or have to be measured.
Beforehand check the definition of H as it differs between the respective books with
respect e.g., to g and mol or bar and kPa.
 5.5 Gas transfer – supplying microorganisms with gaseous compounds | 119

Tab. 5.4: Henry coefficients for different gases and solvents at 298 K (≈ 25 °C).

Henry coefficient Value [g ⋅ L−1 ⋅ bar−1 ] Major dependencies Comments

H O2,H2O 41 ⋅ 10−3 T , salts, surfactants Often limiting
H O2,H2O 35 °C 34 ⋅ 10−3 Decreasing with increasing
temperature
H CO2,H2O 1.48 pH Plus dissociates carbonate
species
H CH4,H2O 22 ⋅ 10−3 Biogas
H H2,H2O 1.54 ⋅ 10−3 Biogas, extremely low
H O2,Decane 570 ⋅ 10−3 Hydrophobic solvent, similarity
to cell membranes
H NH3,H20 490 pH, T strongly Undesired stripping during
decreasing sterilization, highest solubility
for gas in water

As an example we can calculate the oxygen concentration in water for normal air as
gas phase at 25 °C.
g mg
cO2,H2O = 41 ⋅ 103 ⋅ 0.21bar = 8.61 (5.21)
L ⋅ bar L
This is a very low value in comparison to other dissolved nutrients leading to multi-
ple adverse consequences. Firstly, the amount of oxygen stored in the medium is so
low that it would be immediately consumed by the microorganisms unless it is not
constantly supplied by aeration. Secondly, the driving force, being the concentration
difference over the water film (see below), is very low leading to a slow transport from
the bubbles into the medium. In fact, this step can be limiting in industrial fermenta-
tions. Thirdly, uptake by the cells has to be fast even at these low concentrations.
The second important law is Fick’s law of diffusion. The flux FDiff,Gas [g ⋅ s−1 ]
through a membrane is given as:

k Diff,Gas ⋅ ∆cGas ⋅ ABub

FDiff,Gas = (5.22)
dMembrane
This equation is motivated by considering that a higher flux is possible for larger mem-
brane areas and lower for increasing membrane thickness. The diffusion coefficient
k Diff,Gas is a material property. The driving force of this mass transport is the concen-
tration difference over the membrane. For a membrane separating a gas phase and
a liquid phase ∆cGas = c∗G,Gas − cG,Liq , assuming that the concentration on the mem-
brane surface at the gas phase side is in equilibrium with the gas phase. That may not
be entirely true, as additional boundary layers exist on both sides of the membrane.
Quasistationary conditions follow a linear progression of the concentration inside the
membrane.
120 | 5 Bioreactors – designing a home for the bioreaction

For application of Henry’s and Fick’s law to gas transport from bubbles to the
medium we imagine a virtual liquid boundary layer (‘film’) of thickness dFilm between
the bubble surface and the free fluid as first achieved by Nernst (1904). Behind this
‘film theory’ stands the observation that close to the bubbles the water layer is laminar
and molecules are somehow ordered. Oxygen molecules can pass this boundary layer
only by molecular diffusion. Gas transfer is therefore diffusion limited. For oxygen that
reads:
k Diff,O2,H2O ⋅ (c∗O2,H2O − cO2,H2O ) ⋅ ABub
FDiff,O2 = (5.23)
dFilm
This equation supports understanding of the gas transport but is not simply applicable
to real situations inside the bioreactor. While the diffusion coefficient k Diff is tabulated,
the thickness of the film dFilm is hardly measured and is not exactly defined. This gives
reason to introduce the gas transfer coefficient k L [m ⋅ s−1 ]:
k Diff
kL = (5.24)
dFilm
Its value can be measured with macroscopic means e.g., for defined plane surfaces.
The total bubble volume related to the fluid volume ϵ = VBub /VLiquid is the gas holdup.
However, in a bioreactor the total bubble surface area ABub is usually not known. The
next step towards a measurable number is to relate the bubble volume to the fluid
volume giving the volumetric exchange area as aBub = ABub /VR . Finally aBub and k L
are multiplied to give the volumetric gas transfer coefficient k L a [1 ⋅ s−1 ], or the ‘k L a
value’ for short. This procedure of combining physical parameters with macroscop-
ically measurable characteristic coefficients is a common approach in process engi-
neering (compare with derivation of enzyme kinetics). The oxygen transfer rate can
now denoted in a compact form as:

OTR = kL a ⋅ (c ∗O2 − c O2 ) (5.25)

Typical values for the above mentioned parameters are given in Table 5.5.
Besides this ‘stagnant film’ model more complex models can be employed consid-
ering a partial exchange of water in the film (‘surface renewal’ model) or assuming an-
other boundary layer inside the gas phase (‘two film’ model). For precise investigation
of the molecular mechanisms involved in mass transfer the diffusion coefficient is re-
placed by the Schmidt number Sc = ν/k diff describing the relation between kinematic
viscosity and molecular diffusivity. This helps to predict values for different gases,
fluids, relative velocities, and temperatures. Nevertheless, for practical applications
equations are condensed as described above and k L and/or k L a have to be experi-
mentally determined. A rough correlation between k L a and aeration rate/agitation is
given in Equation (5.26):
PR a
k L a = A ⋅ ( ) ⋅ UGb (5.26)
VR
 5.5 Gas transfer – supplying microorganisms with gaseous compounds | 121

Tab. 5.5: Typical parameter values describing gas transfer.

Variable Abbreviation and unit Values for Comments

oxygen
Diffusion coefficient kDiff,O2 [cm2 ⋅ s−1 ] 2.42 ⋅ 10−5 CO2 2.1; N2 2.0;
[25 °C]
Film thickness d Film [m] 0.5⋅10−3 –1⋅10−4 Increasing with velocity
and turbulence of phases
Gas transfer coefficient kL,O2 [m ⋅ s−1 ] 8 ⋅ 10−4 –3 ⋅ 10−4 For d Bub > 2 mm
1.5⋅10−4 –8⋅10−4 for d Bub < 2 mm,
increasing size
Bubble ascent velocity v Bub [m ⋅ s−1 ] 0.2–0.3 For d Bub > 2 mm
0.02–0.3 for d Bub < 2 mm
Volumetric surface a
aBub [m2 ⋅ m−3 ] 1200–120 For d Bub = 1–10 mm,
exchange area ϵGas = 20%
Gas holdup ϵGas [Vol.%] 10–40 aBub = 6 ⋅ ϵGas ⋅ d −1
Bub for
spheres
Volumetric gas transfer kL a [1 ⋅ s−1 ] 0.04–0.4 Depends on surface active
coefficient compounds
Volumetric oxygen OTR [g ⋅ L−1 ⋅ h−1 ] <5
transfer rate

Where A, a, and b are empirical parameters. As other correlations also exist great
caution is necessary. The best way is to ask the supplier in case a new reactor shall be
purchased or to rely on measurements.
For bubble columns it is mainly determined from bubble diameter dB and the gas
holdup ϵR . The k L value is nearly constant ignoring changes in film thickness due to
different rising velocities.

kL 6g kL
kL a = ⋅ dB ⋅ = 6⋅ ⋅ εG (5.27)
dB dB dB

This holds at last for the laminar regime (compare with the microalgae in Chapter 8),
where the bubble diameter can be adjusted by the sparger. A commonly applied for-
mula for the general case of bubble columns is:

k L a = 0.467 ⋅ UG0.82 (5.28)

The background is that bubble diameter no longer depends on the sparger but on tur-
bulences and can therefore no longer be measured.
A first impression of how different aeration and agitation rates look in reality are
depicted in Figure 5.21.
How much oxygen do cells need to grow? The oxygen demand is quantified by the
oxygen yield coefficient Y X,O2 = ∆m X /∆mS similar to Y X,S for the substrate. A typical
122 | 5 Bioreactors – designing a home for the bioreaction

0.05 vvm; 0 rpm 0.05 vvm; 500 rpm 0.05 vvm; 750 rpm 0.05 vvm; 1000 rpm

0.5 vvm; 0 rpm 0.5 vvm; 500 rpm 0.5 vvm; 750 rpm 0.5 vvm; 1000 rpm

Fig. 5.21: Pictures for different aeration rates (vvm) and agitation speeds (rpm).

value is Y X,O2 ≈ 1 g ⋅ g−1 for aerobic growth on glucose. This value can be understood
making the rough assumption that for Y X,S = 0.5 g ⋅ g−1 half of the glucose is used to
build up biomass and the other half is oxidized in a molar ratio of 6 mol O2 /1 mol glu-
cose corresponding to 192/180 g ⋅ g−1 ≈ 1 g ⋅ g−1 . On the level of the specific turnover
rates we obtain:
OUR = rO2 ⋅ c x (5.29)
As glucose uptake, growth rate, and oxygen uptake are stoichiometrically coupled, rO2
depends on the limitation conditions. The specific oxygen uptake rate itself is assumed
to be kinetically limited at partial pressures below 10% or even 5%.
Carbon dioxide production is described by analogy as:
CPR = rCO2 ⋅ c x (5.30)
Assuming that oxygen is almost purely used in respiration and carbon dioxide mainly
produced there, it makes sense to look at the stoichiometric relation between carbon
dioxide production and oxygen consumption, the respiratory quotient RQ:
CPRmol
RQ = (5.31)
OURmol
RQ of 1 mol/mol indicates aerobic growth on glucose. Formation of side products like
ethanol in the case of yeasts leads to higher values as CO2 is formed in the anaerobic
 5.5 Gas transfer – supplying microorganisms with gaseous compounds | 123

pathway. Growth on substrates with higher or lower degrees of reduction than glucose
leads consequently to changes in the RQ value as well. So this parameter is a useful
measure to follow changes in the metabolic pattern of fermentations.

5.5.2 Measuring gas transfer

Determination of OTR und CTR is possible via balances over in-gas and out-gas. In the
following equations the ideal gas law is employed. The definition of FGas refers to stan-
dard conditions TGas = 273.15 K, PGas = 1.013⋅105 N ⋅ m−2 and the universal gas con-
stant R = 8.314 J ⋅ mol−1 ⋅ K−1 . The molar volume Vm = R ⋅ TGas /PGas = 22.4 L ⋅ mol−1 .
In the first step we determine the volumetric molar gas flow to the inlet:
fGas,in ⋅ PGas
FGas,mol,in = (5.32)
R ⋅ TGas
and to the outlet accordingly:
fGas,out ⋅ PGas
FGas,mol,out = (5.33)
R ⋅ TGas
Note that the volume flow into and out of the reactor is not necessarily the same. Why?
In the second step the volumetric mass flows for the different gas compounds,
here O2 , ensues as
fO2,mass,in = fO2,mol,in ⋅ MO2 ⋅ xO2,in (5.34)
and for the other gas species and the exhaust gas respectively.
Usually we know the gas composition of air beforehand xO2 = 0.2095, xCO2 =
0.0004, xN2 = 0.7901 (inclusive other inert gases), but that has to be checked de-
pending on the gas origin like bottles with artificial air or the space from which the
air is pumped. The composition of the exhaust gas is obtained from measurements. In
any case the total molar balance holds for the exhaust gas:
xO2 + xCO2 + xinert = 1 (5.35)
The inert gases (mainly N2 ) leave the reactor, as the name already suggests, unchanged
with respect to moles and mass:
fN2,in = fN2,out (5.36)
We can now derive in the third step the formula for the volumetric transfer rates:
OTR = fO2,mass,in − fO2,mass,out
fO2,mol,in ⋅ MO2 ⋅ xO2,in ⋅ (1 − xCO2,out ) − xO2,out ⋅ (1 − xCO2,in )
= (5.37)
1 − xO2,out − xCO2,out
CTR = fCO2,mass,out − fCO2,mass,in
fCO2,mol,in ⋅ MCO2 ⋅ xCO2,in ⋅ (1 − xCO2,out ) − xO2,out ⋅ (1 − xCO2,in )
= (5.38)
1 − xO2,out − xCO2,out
124 | 5 Bioreactors – designing a home for the bioreaction

cO2 [g·l-1]
Difference
measures
OTR/kLa
1/kLa
c*O2
Exponenal
increase
cO2,meas

63.2%
Linear decrease
Slope measures -OUR

Aeraon off Aeraon on Measuring

OTR = 0 me tmeas [s]

Fig. 5.22: Course of oxygen concentration during an experiment to determine OTR, kL a, and c ∗O2
during an ongoing cultivation.

Note that OTR is defined with a reverse sign than CTR to get positive values for the
normal operational conditions. In cases where more CO2 is being produced than O2 is
being consumed (RQ > 1) the exhaust gas flow is greater than the inlet flow.
Now we make use of the measured OTR value to determine the k L a value from
Equation (5.39) as:
OTR
kL a = ∗ (5.39)
cO2 − cO2

In not all processes, particularly lab reactors, are off-gas measurements pro-
vided, as the off-gas analyzer is expensive. An estimation can be obtained by the so
called dynamic k L a measurement or ‘gassing in – gassing out’ method. The approach
can be carried out in a running fermentation starting from steady state conditions
(OTR = OUR) as shown in Figure 5.22. The saturation concentration c∗O2,ex has to be
determined for the gas phase composition in the reactor, practically the exhaust gas.
In fact, this value is not known without off-gas measurement. To start the measure-
ment, aeration is stopped (OTR = 0). A linear decrease of oxygen concentration ac-
cording to dcO2 /dt = OUR can be observed and determined from the measurements.
Switching on aeration brings the system again to steady state, where cO2 follows an
exponential curve (5.40) with the formal solution of the differential equation:
OUR
cO2 (t) = (c∗O2 − ) ⋅ (1 − e−kL a⋅t ) (5.40)
kL a
We know already the oxygen concentration for the steady state condition, which al-
lows us to substitute OUR/k L a by (c∗O2 − cO2,meas ) and make further simplification by
cancelling the unknown term k L a ⋅ c∗O2,ex leading to:

cO2 (t) = cO2,meas ⋅ (1 − ekL a⋅t ) (5.41)

 5.5 Gas transfer – supplying microorganisms with gaseous compounds | 125

It is not necessary to wait until equilibrium is obtained. For taer,on = 1/k L a the curve
reaches a value of 63.2% of its way from the minimum to the already measured steady
state (1 − e−1 = 0.632). Another way is to plot the values on a logarithmic scale and
read out the slope. This dynamic measurement is not very precise as e.g., if the sen-
sor is not fast enough to resolve the curve precisely or in cases where a considerable
headspace aeration occurs. Furthermore, cO2,ex changes during the oxygen increase
phase of the experiment. These effects could be considered by logging the full dataset
and estimating the parameters in a computer program including time constants of the
side influences. Small steps of the operational pressure can also be applied, which
overcomes the inaccuracies caused by residual bubbles after switching off of aeration
(dynamic pressure method, DPM).

5.5.3 Alternative means of gas supply

Bubbling is the most commonly used means of oxygen supply for aerobic microorgan-
isms. Aeration by gassing can easily be applied and besides they contribute to mix-
ing. However, there are some concerns. Some sensitive cells like animal cells can be
harmed by surface tension of the bubbles. Floatation leads to attachment of cells to
especially small bubbles. Rising up and bursting at the surface can even destroy cells.
From an engineering view, increasing gas transfer is at the cost of higher gas holdup
and shear stress, which is not really welcome. For special cases other forms of gas
supply have to be found.
One obvious solution is surface aeration like in shaking flasks, where oxygen dif-
fuses from the headspace into the liquid. Shaking, besides providing mixing of the
culture, increases the surface area for better mass transfer. This approach depends on
a large surface to volume ratio, which is the case in shaking flasks due to the rela-
tively thin liquid layer. Furthermore, cell concentration and specific growth rate and
therefore volumetric oxygen demand are usually lower compared to bioreactors.
Employment of membranes is a means to separate the gas and liquid phase and
to give additional degrees of freedom for geometric design of larger bioreactors. As an
example a membrane reactor on the pilot scale is depicted in Figure 5.23.
Such membrane reactors are commonly employed e.g., in animal cell culture.
In living nature bubbling is an absolute exception. So it is worth looking at how
plants and animals solve the transport problem of bringing oxygen or carbon dioxide
from air or water to the cells. Leaves use stomata to provide an unimpeded means of
air transport near to the cells. Chloroplasts are often arranged close to the cell wall to
minimize the diffusion path to a few µm. In fact, CO2 transport is often a limiting factor
for growth. C4 plants like maize can bind CO2 during the night to make profit from
additional hours for gas transport. A technical solution in greenhouses is CO2 gassing.
Some plants also manage air transport over longer distances. The most remarkable
examples are reed and alder. They can grow with their roots underwater or can at
126 | 5 Bioreactors – designing a home for the bioreaction

Fig. 5.23: Picture of a membrane reactor for plant and animal

cell culture. The membrane tubes are wrapped around a pivoted
upper and lower frame. Thus, the membranes themselves form
the agitator (© B. Frahm, Bayer Ag).

least survive longer periods of flooding. Oxygen supply to the roots is possible by so
called aerenchym and internal aeration channels. This feature is technically used in
artificial wetlands for wastewater purification. Animals with a high oxygen demand
are facing comparable problems. Most of the vertebrates need lungs or gills to provide
a large surface area (200 m2 in humans) for diffusion, the first step of oxygen uptake.
As in bioreactors, convective transport is the next step. Comparing solubility of oxygen
in water and blood (up to 70 fold higher in blood) reveals that nature invented a great
solution for this problem by binding O2 to hemoglobin. The last step from the blood
to the tissue again is diffusion. This requires a minimization of the diffusion length by
a network of blood capillaries.
 5.6 Exercises, questions and suggestions | 127

5.6 Exercises, questions and suggestions

1. Usually it is considered that the diffusion resistance of oxygen from the

medium to the cells is negligible compared to oxygen diffusion out of the bub-
bles. What is the volumetric surface area of cells assuming c X = 10 g ⋅ L−1 (dry
weight), cells are spheres of dCell = 2 μm diameter, and the water content of
the cells is 90%?
2. In a cultivation with E. coli we measure cO2 = 5 mg ⋅ L−1 , c X = 10 g ⋅ L−1 and
r X = 0.5 g ⋅ g−1 ⋅h−1 . Suddenly aeration stops. Calculate the time tcrit until the
dissolved oxygen is exhausted.
3. Make a list of the physical parameters determining OTR and discuss qualita-
tively how they can be influenced by technical control parameters. What are
the limitations of manipulating OTR?
4. During a cultivation a nonlimiting oxygen concentration of cO2,low =
1 mg ⋅ L−1 is reached with OTR = 0.1 g ⋅ L−1 ⋅ h−1 . As it tends to become even
lower we decide to increase the stirrer speed. Luckily the new steady state
stabilizes at cO2,high = 2 mg ⋅ L−1 . During the transient increase we measure
cO2,trans = 1.63 mg ⋅ L−1 after ttrans = 100 s. How much changed is OUR, OTR,
c∗O2 , and k L a?
5. You decided to improve the gas transfer in the reactor by changing the
sparger. You have the choice between three different types producing gas bub-
bles with following sizes: Sparger A aBub = 0.3 mm, B aBub = 3 mm, or C
aBub = 10 mm. In comparison the actual sparger produces gas bubbles with
aBub = 1 mm. It will be tenable to change the actual one and in that case,
which sparger would you choose? Base the calculations on the time needed
until 50% of the O2 from the gas bubble diffuses into the medium. Further
data: cL = 0 kg ⋅ m−3 , c∗ = 0.0036 ⋅ cO2,Bub kg ⋅ m−3 ,

4
VBub = ⋅ π ⋅ r3Bub , ABub = π ⋅ r2Bub
3
6. The old sparger will be installed in a new air lift reactor. How high can the
air lift reactor be if 20% of the oxygen from the bubbles is expected to be dis-
solved in the medium before the bubbles leave the reactor?
128 | 5 Bioreactors – designing a home for the bioreaction

1. v X ≈ 100 mL ⋅ L−1 ; ACell /VCell = π ⋅ d2 /(1/6 ⋅ π ⋅ d3Cell ) = 6/dCell = 3 ⋅ 106 m2 /m3 ;

a X = ACell /VCell ⋅ v X = 3 ⋅ 102 m2 ⋅ L−1 = 3 ⋅ 105 m2 ⋅ m−3 . That is three orders of
magnitude higher than the bubble surface area, but do not forget that the driving
force is much lower.
2. rO2 ≈ 0.5 g ⋅ g−1 ⋅ h−1 ; OUR = rO2 ⋅ c X = 5 g ⋅ L−1 ⋅ h−1 ; dcO2 /dt = −OUR;
tcrit = cO2 /OUR = 3.6 s. In such an intense cultivation the dynamic measurement
of k L a is practically not feasible.
3. k Diff,O2 is a basic physical constant, so hardly can be changed. Higher viscosity
means even lower diffusivity. Polysaccharides are a candidate for high viscosity.
k Diff is reduced up to 50%, which is also an interesting value to understand diffu-
sion in biofilms. aBub : increasing gas holdup and reducing bubble size by higher
agitation and stirrer speed. Drawbacks: higher gas volume in the reactor reduces
active volume, small bubbles do not leave the reactor fast enough after oxygen
has been released, and shear stress for the cells. k L,O2 : reducing thickness of the
water film around the bubble by higher agitation; drawbacks the same as above.
c∗O2 : according to Henry’s law increasing of oxygen content of aeration gas, which
may be too expensive for many products; increasing pressure in the reactor, which
is the case for high bubble columns, but needs thicker reactor walls, which is
again expensive. Increasing Henry coefficient by changing chemical composition
of the medium, possibly reduces unnecessary salts concentrations. Larger effects
are seen by mixing in of hydrophobic organic fluids, which has already been tried
with positive results. Probably also the boundary layer was straddled. However,
side effects prevent frequent use.
4. OUR did not change because respiration does not depend on the concentration
for nonlimiting conditions. That also means OTR remains constant. FGas,in was
not changed, so xO2,off did not change and therefore neither did c∗O2 . The new k L a
value is 1/ttrans = 0.01 s−1 . c∗O2 = OTR/k L anew + cO2,new = 4.78 mg ⋅ L−1 . k L aold =
OTR/(c∗O2 − cO2,old ) = 0.00735 s−1 .
5. VBub ⋅ dcO2 /dt = k L ⋅ A(c∗O2 − cO2 ); dcO2 /cO2 = 0.036 ⋅ 6 ⋅ k L /dBub ⋅ dt; Integration
ln(cO2 /cO2,0 ) = 0.036 ⋅ 6 ⋅ k L ⋅ t/dBub , cO2 /cO2,0 = 0.5, t = 0.693 ⋅ dBub /0.036 ⋅
6 ⋅ k L . Taking into account that k L depends on the bubble size k L,0.3 mm = 1.1 ⋅
10−4 m ⋅ s−1 , k L,1 mm = 1.1 ⋅ 10−4 m ⋅ s−1 , k L,3 mm = 5 ⋅ 10−4 m ⋅ s−1 , k L,10 mm =
3 ⋅ 10−4 m ⋅ s−1 → t0.3 mm = 8.72 s, t1 mm = 29 s, t3 mm = 19.2 s, t10 mm = 106 s.
The sparger with aBub = 0.3 mm allows a faster gas transfer than the old one and
will be choose.

6. Known values are: vBub = 10 cm ⋅ s−1 , k L,1 mm = 1.1 ⋅ 10−4 m ⋅ s−1 , dBub = 10−3 m:
ln(cO2 /cO2,0 ) = 0.036 ⋅ 6 ⋅ k L ⋅ t/dBub , ln(cO2 /cO2,0 ) = −0.223, t = 9.39 s, with
vBub = 10 cm ⋅ s−1 the height of the air lift reactor should be 80 cm.
6 Not always so simple – the batch process
reconsidered
Batch processes seem to be quite simple at the first glance: inoculation of the reactor,
doing what we like to do namely nothing, and finally harvesting. However, there are
some hidden pitfalls, which have to be considered beforehand. This chapter is dedi-
cated to description of some possible complications. These include switching between
different limitation and inhibition conditions. To understand the interplay between
the cells and the different medium compounds the topic of ‘kinetics’ also has to be
reconsidered, as well as its influence on the course of a batch fermentation. Finally,
this should lead to rational process development.

6.1 Formal kinetics – extrapolation from enzymatic reactions to

cell growth

Complex reaction systems like the growth of cells in bioreactors have to be described
using strong simplifications. These can be educated guesses about the major steps and
substances that reduce the system description accordingly. Other assumptions can be
gained from global constraints like balances or thermodynamics of the system. We
did already do this when assuming that only substrate uptake is decisive, and growth
follows stoichiometric and thermodynamic constraints. This approach has been tried
for decades in order to find easy manageable formulas for growth in batch and other
process policies. The most important ones are given in Table 6.1. Nevertheless, for ra-
tional process development it is important to understand the assumed simplifications
or observations. These can have their origin on the cell level or on the reactor level. In
practice, not all compounds produced by the cells and accumulating in the reactor in
small amounts are quantitatively detected or the medium is not ideally mixed. Neither
are all metabolic pathways or intracellular control loops. That is why the notations are
called ‘formal kinetics’ They are a practical start to process analysis in the sense of a
modular construction kit but include so many assumptions and simplifications that
conclusions on underlying mechanisms are not allowed. Often they are valid only for
the process type for which they have been formulated, e.g., batch processes. In every
case simultaneous inspection of growth, substrate uptake, and product formation on
demanding is indicated. To go a step further into a more mechanistic view, we have
first to discuss what could eventually happen in a batch process demanding our at-
tention.
Most of the kinetics are deduced from enzyme kinetics for substrate uptake as
in the case of Monod kinetics. This has to be done with great care. The involved en-
zymes are often not characterized in their microenvironment with respect to limita-

https://doi.org/10.1515/9783110315394-006
Tab. 6.1: A collection of commonly applied formal kinetics to describe growth.

Author Equation Rationale behind it

cS
Monod μ = μmax ⋅ kS +cS
Enzyme kinetics
cS
Andrews, Haldane μ = μmax ⋅ c2
Enzyme, substrate inhibition
kS +cS + k S
l,S
cSn
Hill, Moser μ = μmax ⋅ kSn +cSn
Multiple binding sites, reaction order
cS
Contois μ = μmax ⋅ kS ⋅c X +cS
Biomass influences itself
dμ
Tessier μ = μmax ⋅ (1 − e−kS cS ) dcS
= kS ⋅ (μmax − μ)
1−p 1−p dμ
Konak μmax − (μmax − μ) = (1 − p) ⋅ kS ⋅ c S dcS
= kS ⋅ (μmax − μ)p Monod for p = 2, Tessier for p = 1

{μmax ⋅ kcS for c S ≤ kS

Blackman μ(c S ) = { S Special case of enzyme cascade
μmax (= kS ⋅ c S,sat ) for c S > kS kS here not the ‘half saturation constant’, c S,sat the concentra-
{
tion for which μmax is reached
cl cl k l
kS ⋅ (1 + kl
) ; c S,denom ⋅ (1 + kl
) ; c S,nom k +c All three formally possible inhibition types, see Chapter 4 (ki-
l l
130 | 6 Not always so simple – the batch process reconsidered

netics)
−c S −c S
kS
Tessier type with substrate inhibition μ = μmax ⋅ (e kI,S − e )
−c S
cS kI,S
Aiba μ = μmax ⋅ kS +cS
⋅e Monod plus Tessier type inhibition
Edwards, Webb and others Mixed terms from enzyme and Tessier type kinetics
 6.1 Formal kinetics – extrapolation from enzymatic reactions to cell growth | 131

tion constants and turnover numbers. Their amounts present in the cell membrane is
unknown. Most microorganisms exhibit more than one substrate uptake mechanism
for different situations like lower or higher substrate concentrations in the medium.
E. coli for example possesses five different glucose transport systems with three dif-
ferent transport mechanisms, and S. cerevisiae has with 17 functional hexose trans-
porters (seven for glucose), the highest number. Nevertheless, a formal kinetics with
two parameters has been found to work reasonably well. The background is that dur-
ing a batch process substrate limitation occurs only during a short time interval, where
only a few measured values are available. A second point is that the additive over-
laying of several Michaelis–Menten kinetics looks quite similar to one with averaged
kinetic parameters. The system is not observable with standard offline measurements.
Substrate inhibition is a clearly defined mechanism in enzyme kinetics. In a batch
processes high substrate concentration has a multivariate impact on the cell and re-
actor levels. This includes a high viscosity with side effects on mass transfer. High
osmotic value is another effect to which the cells have to react by energy expendi-
ture for water pumping or buildup of intracellular osmotic compounds. With respect
to batch processes a long lag phase and slowly reduced inhibition effects during cul-
tivation are the consequences, including changes in different growth characteristics.
Thinking of substrates other than glucose, toxic effects may occur with multiple effects
on the cells. The same holds for product inhibition, where approaches usually known
from enzyme kinetics are applied. Extracellular compounds may not only react with
substrate uptake but intracellular steps may also attack the cell membrane as in the
case of ethanol.
Contois kinetics looks a bit strange at first glance as biomass concentration ap-
pears on the right side of the equations. This means that the cells influence each other.
That could actually be so in cases of quorum sensing, but often a hidden material
transport limitation is the reason that the cells experience a lower local substrate con-
centration than is macroscopically measured.
The kinetics according to Konak with the special case given by Tessier can be inter-
preted as an approach where the difference of the actual growth rate and the maximum
growth rate acts a driving force. The further the organisms are away from the optimum
the more effectively they can use additional substrate. Although the biological back-
ground is quite unclear the kinetics is frequently used because of its variability. For
p = 1 (Tessier) the kinetics is formally identical to the step response of a dynamic sys-
tem with one time constant. For p = 2 the Monod equation is a formal solution. Both
curves look very similar and can hardly be distinguished under practical conditions.
Substrate uptake is not always the limiting step over the whole range of substrate
concentrations. Other steps further down the catabolic pathways can be limiting be-
fore the uptake is at its maximum. This case has been approximated by Blackman as
a piecewise constant kinetic function. The linearly increasing behavior is an approxi-
mation of the Monod kinetics for low substrate concentration, where the maximum
represents a subsequent limiting step, which is masked for low substrate turnover
132 | 6 Not always so simple – the batch process reconsidered

rates. Indeed, Blackman kinetics fits better to the original data of Monod than the
Monod kinetics itself. Apparent low k S values, especially those lower than for the iso-
lated enzymes, can be the effect of high overexpression of the enzymes involved in the
transport system to allow the cell high turnover rates at low substrate concentrations.
Nevertheless, the apparent μ max is not given by the maximum of the transport system
but by another intra- or extracellular limitation.
Other kinetics are valid for special cases, e.g., when an unknown limitation or
inhibition is active during a special process pattern. Here our job is not simply to try
out different formal approaches but to find out the mechanics behind it and set up
equations for their precise representation. Before trying it out for our standard batch
good guesses for reasonable values of the yield parameter y X,S are required.

6.2 Looking a step deeper – estimating aerobic growth yields

from metabolic fluxes

Microbial metabolism is subject to the energy balance represented by ATP and the
redox balance represented by NADH2 . Remembering Liebig’s barrel model, ATP and
NADH2 can be visualized as the hoops keeping all the staves together. A more quanti-
tative approach to find good values for yields as a link between substrate uptake and
growth starts with evaluation of the main metabolic pathways as shown in Figure 6.1
for aerobic growth on glucose.

O2
2 · nC6 6 · nC1

Glucose TCC CO2

nC6 nC1 H2O

2 · nC6
6 · nC1 6 · nC1 ATP ADP

NADH NAD
nC1
Biomass

Fig. 6.1: Simple metabolic pathways diagram of aerobic growth for estimation of ATP and NADH2
balance. The metabolic pathways are given in molar fluxes q for the C skeleton with carbon atoms
indicated as Cn in the index. Assumptions: 2 NADH2 , 2ATP per C6 in glycolysis, 1 NADH2 per C3 pyru-
vate oxidation, 3 NADH2 + 1QH2 + 1 ATP per C3 acetyl oxidation in TCC, 1 NADH2 → 10 H+ membrane,
1 QH2 → 6 H+ membrane, 4 H+ membrane → 1 ATP.
 6.2 Looking a step deeper – estimating aerobic growth yields from metabolic fluxes | 133

With the given simplifications we can set up the balance equations:

ATP-balance:
2 ⋅ qglyc + 15 ⋅ qresp − 5 ⋅ q X = 0 (6.1)
NADH2 -balance:
1
2 ⋅ qglyc − (6 − 5) ⋅ qresp − ⋅ qX = 0 (6.2)
3
Normalizing qglyc to 1 the solution is:

2 ⋅ 16
qX = ≈3; qresp ≈ 1 (6.3)
5 ⋅ 14 − 5 ⋅ 16
With the given simplifications it can be fixed that glucose can be completely oxidized
yielding 32 ATP per mole. Glucose going into biomass utilizes 30(6 ⋅ 5) mole ATP per
mole glucose. From these relations follows a carbon flux allocation of about qresp,C :
q X,C = 1 ⋅ 3 : 3 ⋅ 1 = 1 : 1 into respiration and into anabolisms. This value is in-
deed sensible, as in many cultivations a yield y X,S ≈ 0.5 g ⋅ g−1 for aerobic growth on
glucose is measured.
For the specific values of ATP and NADH2 production in the diagram, growth and
respiration are only coupled by ATP as the degree of reduction of biomass is the same
as the one of glucose. In principle, all glucose can be respired under ATP production
without violating the redox balance. The same holds for allocation of glucose into
growth provided there is another ATP source. Changes in the degree of reductions of
biomass would change the situation in so far as the redox balance couples growth and
respiration, making the oxygen consumption no longer follow a molar ratio of 1:1 to
carbon dioxide production. Why we employed the ATP balance strictly, and did not
consider something like decoupled ATPases, is an implicit suggestion of an internal
mechanism of the cell that comes as close to optimum conditions as possible. In fact,
that is not always the case.
The deduction of real yields from the values given in the metabolic pathway has
to be done with some care. The P/O ratio (mole ATP per mole O) is implicitly given here
as 2.5 (15/6). Membrane leakage can lead to loss of protons necessary to maintain the
mitochondrial proton gradient driving ATP formation. In references a practical value
of 1.75 mole/mole is often assumed. Similarly, uncertainty has to be faced for ATP de-
mand fueling anabolic pathways for cell growth. This is e.g., given as 100 mmole ATP
per g biomass (calculation from Figure 6.1 yields in 200 mmole/g). The lower estima-
tions are based on the assumption of a lower P/O ratio. So the ratio, which we are
basically are interested in, stays constant. Some attempts have been undertaken to
account for the number of ATP molecules considering all known metabolic pathways
from the synthesis of the metabolites up to the macromolecules. This gives only a part
of the real demand. More growth related energy is necessary for transport processes,
repair and other diffuse losses. Difficulties in interpretation of literature data comes
from lumping too different reference points or allocation of losses to different path-
ways. Another point is biomass composition. Here it is assumed that the degree of
134 | 6 Not always so simple – the batch process reconsidered

reduction of biomass is the same as that of glucose. In fact, real values are lower e.g.,
due to reduced compounds like fatty acids. Further, we neglected de-/carboxylation
and de-/hydration processes distributed over many sites in the metabolism. More re-
ductants given as NADH2 equivalent are necessary to build up reduced compounds
on the one hand but are in that case not available in the respiratory chain on the other
hand. This leads to a reduced growth yield on a biomass basis. Nevertheless, lipids
have a higher energy content therefore reducing the losses in yield on an energy basis.

6.3 Aerobic growth – case study of heterotrophic plant cells

Plant cells can be cultivated under aerobic conditions on sucrose as substrate. This
actually happens in growing plants where the cells not involved in photosynthesis
are supplied with sucrose from the leaves. Cultivation data are shown in Figure 6.2.
After a lag phase of about one day the cells start to grow with a specific growth rate of
about 0.21 d-1 . Interestingly, sucrose is not taken up directly but cleaves to glucose and
fructose due to a strong invertase activity provided to the cells. This reaction is over
after two days. Both monosaccharides can be taken up, while glucose is obviously the
preferred substrate and is consumed first. Growth on two different substrates with dif-
ferent preferences is a common phenomenon during batch cultivations and is called
‘diauxic’ growth (‘aux-’, Latin auxilium = help).
The reasons for preferences can be more active uptake systems, different yield
coefficients meaning more ATP or carbon for the cells, or different catabolic pathways
meaning different expenditure with respect to necessary enzyme makeup. In many
cases the physiological reasons are not clear. Fructose is taken up by fructokinase,
while F6P enters glycolysis directly after G6P isomerase from glucose uptake. Thus,
the energetic yield is the same for both sugars and there is even one enzymatic step less
for fructose in comparison to glucose uptake. In wine making sucrose is hydrolyzed to
glucose and fructose, which are fermented to ethanol with a preference for glucose.
This existing biological preference is under investigation since this mechanism often
leads to incomplete fructose fermentation during wine production, thus affecting wine
flavor quality.
To find a way to describe the concept of a preferred substrate we assume that glu-
cose is taken up following a Michaelis–Menten kinetics:
cGlu
rGlu = rGlu,max ⋅ (6.4)
k Glu + cGlu
Fructose is taken up only in amounts to fill up the gap of the hexose flux induced by
glucose limitation. For fructose uptake in addition a limitation from low concentra-
tions also has to be considered:
cFru
{rFru,max ⋅ for rFru ≤ rGlu,max − rGlu
rFru = { k Fru + cFru (6.5)
{ r Glu,max − r Glu else
 6.3 Aerobic growth – case study of heterotrophic plant cells | 135

30

25

20
c X , c s , c f [g/L]

15 cX
cFructose
10
cGlucose
5 cSucrose

0
0 2 4 6 8 10 12 14 16

(a) Culvaon me [days]

2.0 14

1.8
12
1.6 cNitrate
1.4 10
cN , cA, cP [g/L]

cx [g/L]

1.2
8
1.0
6
0.8 c Biomass
0.6 cAmmonia 4

0.4
2
0.2 cPhos
0.0 0
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
(b) Culvaon me [days]

Fig. 6.2: Measurements of a batch culture with plant cells (Echinacea); continuous lines are simula-
tions. (a) Biomass and main substrates, (b) mineral nutrients.

For the specific growth rate we get:

μ = y X,Glu ⋅ rGlu + y X,Fru ⋅ rFru (6.6)

As expected, both yield coefficients have the same numerical value in this case as both
hexoses deliver the same amount of ATP and carbon.
In this plant cell example there are two nitrogen sources as well, namely ammonia
and nitrate, Figure 6.2. Here the same phenomenon of a preferred substrate can be ob-
136 | 6 Not always so simple – the batch process reconsidered

Tab. 6.2: Three formal kinetics for several substrates and further structured approaches.

Referenced as Formal Applicable for Structured

interactive, μ = μmax ⋅ ∏i r i (c S,i ) Stoichiometrically μ = min(y X,S,i ⋅ rS,i,pot ) ;
multiplicative coupled essential rS,i = y X,S,i
1
⋅μ
substrates
additive μ = μmax ⋅ 1
i
⋅ Alternative substrates To be modified by
∑i r i (c S,i ) without preference physiological control
noninteractive μ = μmax ⋅ r1 (c 1 ) or Alternative substrates considerations e.g., one
μmax ⋅ r2 (c 2 ) with strong prefer- limitation and
ence replenishment to maximum
coupled Stoichiometrically One limiting, one on demand
dependent

served. Ammonia is taken up on demand (nitrogen fraction of the cell), while nitrate
is only used if ammonia uptake is reduced by limitation. This can be motivated by the
necessity of using NADH for reduction of NO−3 . Phosphate is taken up much faster than
expected to cover the P demand of the cell. Phosphate can be stored in the large vac-
uole of plant cells. To store minerals in different chemical forms is a common behavior
of microbial cells. For process development that makes calculation of ideally balanced
media difficult. Giving the total amount of e.g., phosphate in the beginning leads to
polyphosphate accumulation in the cell, but it is not certain whether this pool can be
remobilized fast enough in a later stage of the culture. Kinetics for several substrates
in their interplay are given in Table 6.2.
Multiplicative kinetics (Tsao, Hanson) is an approximation for the case of stoichio-
metrically coupled substrates. Examples are oxygen versus substrate uptake or nitro-
gen and phosphate uptake versus growth. In these cases only one substrate can be
limiting while the other is taken up on demand according to the cell’s elemental com-
position or other intracellular stoichiometries. We already made use of this microbial
behavior for media design. The approximation holds as in many practical applications
only one substrate is in limiting concentrations, while the others may accumulate to
saturation concentrations shifting the value of the respective Michaelis–Menten term
to one. Nevertheless, the more structured formulation should be preferred.
The additive approach is the most reasonable one, as the cells try to take up both
substrates as long as they can further metabolize them.
The noninteractive approach (no happy phrase) stresses the concept of a preferred
substrate. Only in rare cases it is a yes or no decision. It is a physiologically interesting
task to find formulations for the intracellular control principle as given in the plant
cell example. In cases where one substrate is absolutely preferred, a lag phase may
occur before the second one is used. In this ‘diauxic lag phase’ the enzymatic set of
the cells is adapted to the new substrate.
6.4 Looking a step deeper – estimating anaerobic product yields from metabolic fluxes | 137

Diffusion Stoich. of oxidaon Stoich. of ATP generaon

rO2 rKat,O2 rATP
rO2

rKat rKat,O2
cO2

One at kinec maximum Completely stoichiometrically

One down-regulated determined part of network
Controller
rS rKat rX

cS rS rAna rAna,ATP

M.-M. kinecs Free distributor (TCC) Balanced need for carbon and energy

Fig. 6.3: Flow chart of the stoichiometry and uptake control for substrate and oxygen.

The case of substrate and oxygen uptake is represented as a flow chart in Figure 6.3.
Oxygen uptake and substrate turnover is stoichiometrically coupled by respiration,
described by strongly coupled kinetics. ATP production in respiration is strongly
coupled to growth. So basically, two growth patterns can be distinguished, namely
substrate limited growth and oxygen limited growth, where the respective other sub-
strate is taken up by demand. This is represented by the switch in the left box.

6.4 Looking a step deeper – estimating anaerobic product yields

from metabolic fluxes

In anaerobic fermentations no oxygen is available as an electron acceptor. NADH2

has to be used for reduction of another external electron acceptor or an intracellu-
lar metabolite. This is called endogenous respiration. The case of pyruvate reduction
and ethanol formation is shown in Figure 6.4.
With the given simplifications we can set up the balance equations:
ATP balance:
2 ⋅ qglyc + 0 ⋅ qform − 5 ⋅ q X = 0 (6.7)
NADH2 balance:
1
2 ⋅ qglyc − 1 ⋅ qform − ⋅ qX = 0 (6.8)
3
Normalizing qglyc to 1 the solution is:

2 2 8
qX = and qform = 2 − = (6.9)
5 15 15
The allocation of carbon to growth is q X,C /qGlyc,C = 2 ⋅ /(6 ⋅ 5) = 1/15 ≈ 0.067.
Measured values for Y X,S are in the range of 0.03–0.05 g g−1 . This value is in good
138 | 6 Not always so simple – the batch process reconsidered

qglyc,C6 2 1 qferm,C3
Glucose (C6) Pyr Ethanol (C2)
2 CO2 (C1)

ATP ADP 1/3 5

NADH NAD
qX,C1
Biomass

Fig. 6.4: Simple metabolic pathways diagram of anaerobic growth under ethanol formation for es-
timation of ATP and NADH2 balance. The metabolic pathways are given in molar fluxes q for the C
skeleton with carbon atoms indicated as Cn in the index. Assumptions: 2 NADH2 , 2ATP per C6 in
glycolysis, 1 NADH2 per C3 pyruvate reduction.

agreement with the ‘theoretical’ value based on the metabolic scheme. Only a small
part of the glucose is allocated to growth. In anaerobic cultivation a low biomass
concentration can be expected. However, that is not a disadvantage as the intention
is to get a high yield YP,S of the end product, being ethanol in the case described
here. q ferm,C/qGlyc,C = 28/15 ∗ 3/6 = 14/15 shows that nearly all carbon of the
substrate is channeled into ethanol and CO2 . For ethanol as given in Figure 6.4
YEth,Glyc = qferm,C /qglyc,C ⋅ MEth /(MCO2 + MEth ) = 14/15 ⋅ 44/90 ≈ 0.46. In terms
of product formation we have the very good result that about 95% of the theoretical
maximum yield of glucose splitting to ethanol and carbon dioxide can be reached.
The remaining part contributes to growth.

6.5 Anaerobic batch culture – case study of ethanol production

Ethanol production for alcoholic beverages is probably the oldest deliberately per-
formed biotechnological process in human history. Ethanol production for fuel, so
called ‘bioethanol’, works biochemically speaking in the same way, by anaerobic fer-
mentation of sugars. Glucose comes in the case of wine from the grapes or other fruits.
In case of starch an enzymatic step has to precede it: amylase is produced by the malts
in case of beer brewing, but must be provided in a separate step for production of
bioethanol.
Before we go into details of ‘know-how’ an engineer should think about ‘know-
why’ with respect to a useful contribution to human needs, or more strictly speaking
to a viable market. Already Henry Ford proposed in the early 1920s: “The fuel of the
future is going to come from fruit like sumach . . . – almost anything.” However, due to
the emerging petrol industry this wish was not realizable at that time. The basic idea
is to transform a diluted or otherwise difficult to manage substrate into a workable
 6.5 Anaerobic batch culture – case study of ethanol production | 139

liquid fuel or drop-in fuel to be directly usable e.g., for vehicles. As we are going to
produce a fuel the energetic efficiency should be the first step to check. This can be
done by calculating the energetic efficiency ηC1,C2 of conversion on the basis of heat
0
of combustion HC,Comp of the respective compounds:

0
HC,eth ⋅ 2 ⋅ mol 1367 kJ ⋅ mol−1 ⋅ 2 ⋅ mol
ηEth,gluc = = = 0.97 (6.10)
0
HC,gluc ⋅ 1 ⋅ mol 2815 kJ ⋅ mol−1 ⋅ 1 ⋅ mol

This is indeed good news, as the energetic efficiency is acceptably high. This is also
the case if the 5% loss by biomass formation is accounted for, leading to η Eth,Gluc =
0.92. Here we calculated with the higher heating value, which presumes a complete
recovery of the heat of all side products being CO2 and H2 O. This will not completely
be the case as during cultivation cooling energy is required while for rectification heat
is necessary.
The second question is directed to raw material availability. Plants as feedstocks
is of course the ultimate renewable resource based on sun energy and carbon dioxide
driving photosynthesis. The most abundant substrates are sugar and starch. Biofuels
based on these substrates are called ‘first generation biofuels’. However, the amount
of these carbon sources necessary to substitute a remarkable part of fossil fuels is so
high that a direct competition with food and feed is obvious as arable land is limited.
Furthermore, the nexus between growing plants like sugar cane or corn and water and
energy demand has to be carefully evaluated. In the course of the development of ‘sec-
ond generation biofuels’ lignocellulosic materials, a waste product from agriculture,
comes into focus. As long as wastes from agriculture or food industry are considered,
the conversion to a fuel seems to be uncritical. But hydrolyzation of lignocellulosics
is much slower and more complicated compared to starch. Many life cycle studies are
available to calculate the environmental footprint of bioethanol production. Accord-
ing to these bioethanol is not completely carbon neutral but about 70% of CO2 released
into the atmosphere can be reduced compared to fossil fuels.
The market itself is highly politically controlled. The Kyoto protocol in particular
demands an increasing substitution of fossil fuels by biofuels. The United States and
the European Union set road fuel targets at 5–10% in 2016 in order to meet these de-
mands. The current production (2016) is about 100 million tons worldwide. This makes
up a market volume of about US$100 billion. Nevertheless, cost issues are quite im-
portant for process design, as bioethanol is a very low value product and still more ex-
pensive than fossil fuels, which is a big challenge for engineering. Even small changes
of yield or energetic efficiency means profit or loss for society.
Not all possible substrates are practically suitable. Molasses for instance needs
water evaporation to come to more than 30% dry matter concentration necessary for
high final product titer. Furthermore, several inhibitors including salts prevent highly
intensive processes. A typical process for bioethanol production from starch contain-
ing grains is shown in Figure 6.5. The process starts with delivering the corn, which
140 | 6 Not always so simple – the batch process reconsidered

CO2
Grains + Water + Yeast Disllaon Recficaon Dehydraon
+ Enzymes + Sugar syrups

Mash

Milling Saccharificaon Fermentaon

Sllage
99.7 vol.-%
Bioethanol

DDGS

Evaporaon Drying Pelleng

Fig. 6.5: Flow sheet of a bioethanol production process from grains to bioethanol and coproducts.

contributes a significant part to production costs. The corn is fed to the intake pit, from
where it is conveyed to the precleaning section and finally to intermediate storage si-
los. After undesired materials such as sand or straw are separated i.e., by ventilation,
dry milling is the next process step, to make the starch accessible by the enzymes.
Optimum and uniform particle size of the ‘flour’ in the range of 2–3 mm is required to
meet the optimum between milling costs and hydrolyzation time. This ‘mash’ prepara-
tion is the third step, namely a biochemical one. During liquefaction in a warm water
mash, meant to reduce viscosity, α-amylase enzymes break up the starch molecules
and gelatinize the broth. The final product is malto-dextrin, which is further converted
in a saccharification step by glucoamylase to oligosaccharides and monomeric sugars,
basically glucose. Free sugars make this process sensitive to bacterial contamination.
Heat sterilization of the grains is practically forbidden to save energy and to prevent
starch denaturation. Hygienic design and precise process control with respect to pH
and temperature can contribute to a solution.
Now the real fermentation process of the ‘sweet mash’ from the presaccharifica-
tion can start. The preculture of the yeast is carried out aerobically on a side stream of
the sweet mash and then transferred to the anaerobic fermentation as inoculum. Fer-
mentation can lead to ethanol concentrations of up to 150 g/L. Reaching a high titer
of ethanol during fermentation is crucial as it reduces the energetic costs of rectifica-
tion. One limitation is the concentration of fermentable material in the flour and the
mash. Another problem is the strong ethanol inhibition on substrate turnover, growth,
and yield. One batch in fermenters (a typical word in industry for bioreactors) up to
2000 m3 lasts about three days. The fermentation system is designed in such a way
 6.5 Anaerobic batch culture – case study of ethanol production | 141

that fermentation vessels can be operated in parallel. Such plants have a high degree
of automation including cleaning by means of a CIP system.
The first task in this downstream operation is to separate a clear mixture of wa-
ter and ethanol from the ‘fermented mash’ (fermentation suspension), where the raw
alcohol is withdrawn over the head of the column. This has to be done at moderate
temperature to prevent the proteins from denaturation and keep them valuable for
feed purposes. Further rectification in a multicolumn system is done to reach ethanol
concentrations close to the azeotrope equilibrium. Finally, the ethanol has to be dried
e.g., by molecular sieve adsorption or pervaporation to meet fuel specifications, an-
other constraint from the market. The remaining de-alcoholized liquid, the ‘stillage’
(‘vinasse’ liquid + grain residues + dead yeast), is pumped out for further processing
including thermal and mechanical processing steps to allow for selling an additional
product: the ‘dried distillers grains with solubles’ (DDGS), which is used as cattle feed.
A modern bioethanol factory including DDGS production is shown in Figure 6.6.
A high degree of heat recirculation between rectification and mash evaporation
and stillage drying is crucial for minimization of energy losses. The attempt to make
use of the whole value of a complex feedstock to obtain several products has found its
expression in the concept of the ‘biorefinery’. This type of process integration, which
is typical for bioprocesses, will be further discussed in Chapter 8 (Microalgae).
To assess bioethanol production via anaerobic fermentation a look around at com-
peting technologies is vital. The increasing amount of electricity retrieved by photo-

Fig. 6.6: Building a modern bioethanol production; in this stage the process equipment is already
mounted including fermenters, rectification columns and dryers (© GEA Wiegand GmbH).
142 | 6 Not always so simple – the batch process reconsidered

voltaics or wind power has brought up the idea of producing hydrogen by electrolysis
from surplus energy. This hydrogen can be further processed on a biotechnological
route by syngas fermentation (Chapter 10) or by chemical processes summarized as
power-to-fuel or power-to-X technologies, where syngas (hydrogen and carbon diox-
ide) is converted to methane as fuel gas or used as educt for a Fischer–Tropsch process
to yield liquid fuels.

6.6 Running example of yeast production – growth in batch

processes

Yeast has to be produced for selling mainly to bakeries, but also for breweries or wine
makers. The first trial of a production process is to do it as simply as possible a reason-
able batch process. Indeed, this has been the standard procedure for centuries. Look-
ing at cultivation data (Figure 6.7), we are surprised that the batch can be divided into
several phases. In the first one, the yeast takes up glucose as expected but produces
ethanol simultaneously despite the fact that enough oxygen has been supplied. A lim-
itation in oxygen mass transfer would lead similarly to ethanol production. It would be
visible by low pO2 values e.g., lower than 10% saturation. Furthermore, an increasing
volumetric oxygen uptake rate OUR proportional to biomass propagation is observed,
which is proof that mass transfer is not critical in this particular experiment. In the
case of oxygen limitation, the simultaneous activation of the aerobic and the anaer-
obic glucose degradation pathway is called the ‘Pasteur effect’. The yeast cells take
up as much oxygen and glucose as they can. Respiration is then limited by oxygen
availability. Many yeasts do not downregulate glucose consumption to fit with oxygen
uptake but use the excess glucose via the anaerobic pathway for ethanol production.
After a while, glucose is exhausted but the yeast keeps on growing after a short diauxic
lag phase. Yeast is obviously able to take up the self-produced ethanol and use it for
growth. This is a completely aerobic process as can be seen by the further increase of
OUR.
To get a more detailed analysis of the data a look at the specific turnover rates of
the batch experiment is informative (Figure 6.7). The data show that the specific oxy-
gen uptake rate keeps constant during growth on glucose. Additionally, the specific
glucose uptake rate rS is nearly constant, presumably close to rS,max . As the substrate
concentration reaches zero, substrate uptake decays to zero as well according to up-
take kinetics. During growth on ethanol both the specific ethanol uptake rate and the
specific oxygen uptake rate slowly rise proportionally. This is a hint at an intracellular
stoichiometry.
From molecular biological analysis it turns out that oxygen uptake as such is not
the limiting step under high oxygen and sugar availability but a limitation in the respi-
ratory chain. The capacity of the respiratory chain is thus limiting complete oxidation
of the available sugar. This is called ‘overshoot metabolism’ or more specifically the
 6.6 Running example of yeast production – growth in batch processes | 143

30 3.0
14 14

25 2.5
12 12

20 10 10 2.0

Biomasse X [g/L]
Glucose S / [g/L]

Ethanol E / [g/L]

8 8

OUR / [g/L.h]
15 1.5

6 6
10 1.0
4 4

5 0.5
2 2

0 0 0 0.0
0 5 10 15 20

(a) Culvaon me T [h]

f

1.0 0.20 0.3 1.0

0.8
0.15
0.2
0.5
0.6
qS [g/g.h]
qO2 [g/gh]

μ [1/h]
rE [g/g.h]

0.10
0.4
0.1

0.0
0.2
0.05

0.0
0.0

-0.5 0.00
0 2 4 6 8 10 12 14 16 18 20
(b) Time [h]

Fig. 6.7: Measurements during a typical yeast batch process. (a) Substrate, biomass, ethanol and
the volumetric oxygen uptake rate; initial substrate concentration is 28 g/L. (b) Calculated specific
turnover rates of the main substrate and products shown in the yeast batch process.

‘Crabtree effect’ (after Herbert Grace Crabtree, 1928). The maximum oxygen turnover
rate is denoted as the critical specific oxygen turnover rate rO2,crit and the correspond-
ing substrate turnover in the aerobic part of metabolism rS,crit . In yeasts the onset of
the overshoot metabolism may happen for glucose concentration higher than about
cS,crit = 180 mg/L, which corresponds to a substrate uptake rS,crit lower than rS,max .
At low substrate turnover rates rS < rS,crit approximately one half ((1 − y X,S ) ⋅ rS ) of
the substrate is oxidized while the other half is used for growth. Surplus of glucose is
allocated to the anaerobic fermentation pathway of course with lower energetic yield.
Assuming a given substrate uptake rate, growth, respiration and ethanol formation is
stoichiometrically coupled to substrate uptake under the constraint of an upper res-
piration limit as shown in Figure 6.8.
144 | 6 Not always so simple – the batch process reconsidered

rX, rO2
Maximum respiraon
capacity
rO2,max

Growth
rS

rS,crit

rEth

Fig. 6.8: Principal course of specific turnover rates as a function of substrate uptake; after reach-
ing the limiting step for respiration carbon is more and more allotted to ethanol yielding in higher
growth rate but with lower yield.

Glucose Glucose Glucose

bottle neck
Model
CO2 CO2

Ethanol
Ethanol

Biomass CO2 H2O Biomass CO2 H2O Biomass CO2 H2O

Aerobic Growth Crabtree-Effect Pasteur-Effect

Fig. 6.9: Graphical bottleneck model of the Crabtre -effect; the blue ring represents the capacity
of respiration, while ethanol formation occurs either by overloading with substrate (Crabtree) or
oxygen deficiency (Pasteur). In the latter case the blue ring is smaller.

Limiting steps in microbial metabolisms are often called a ‘bottleneck’, resembling a

bottle where pouring and drinking is limited by the bottleneck. A graphical represen-
tation of the bottleneck model is depicted in Figure 6.9.
The rise of the specific oxygen uptake rate during growth on ethanol shows clearly
that the yeast cells might well be able to increase respiratory activity. That the cells do
not do it during growth on glucose is thought to be due to an intracellular optimiza-
tion algorithm. On the one hand, keeping the activity of the respiratory chain on a
high level is a metabolic burden for the cell. On the other hand, the cells try to grow
as fast as possible. This is finally possible at a specific respiration rate but at the cost
 6.7 Primary metabolites – products directly taken from central metabolism | 145

of lost carbon in the form of ethanol. So, respiration is reduced in the presence of glu-
cose, a phenomenon known as ‘catabolite repression’. In the absence of glucose res-
piration is upregulated again, because there is no alternative. Catabolite repression
is also the mechanism for choosing preferred substrates. The best known example is
the preference of glucose over lactose in E. coli and may also play a role in the pant
cell example. We will come back to this optimization aspect in the modeling chap-
ter. Overshoot metabolism against the background of intracellular limiting steps is a
common phenomenon in several bioprocesses, e.g., E. coli produces acetate at high
glucose concentration (sometimes called the ‘bacterial Crabtree effect’).
The Crabtree effect as a characteristic of yeast metabolism has obvious disadvan-
tages in batch processes. Firstly, the overall yield is lower than expected as the cell
cannot make up the low yield from the first growth phase during the second growth
phase. Secondly, accumulation of ethanol – even to higher values compared to the
example in the case of higher initial glucose concentration – leads to ethanol inhibi-
tion. Thus, an efficient yeast production in batch processes is prevented. In the next
chapter we will see what to do in such cases.

6.7 Primary metabolites – products directly taken from central

metabolism

We have already seen that product formation during anaerobic growth is directly cou-
pled to energy metabolism. In aerobic processes product formation may be directly
coupled to catabolic and/or anabolic pathways as well. Excretion of primary metabo-
lites often has its kinetic reason in intracellularly limiting steps, either naturally or
intentionally induced by genetic engineering. These are examples where substrate up-
take limitation and a corresponding yield is not sufficient to describe growth kinetics,
but in addition the assumption of another limiting intracellular step and a correspond-
ing product formation is necessary. A general structure in terms of metabolic fluxes is
given in Figure 6.10.
Intentionally applied oxygen limitation in analogy to the Pasteur effect is ap-
plied in microaerobic processes like 1,3-propandiol production. This allows the cells
to gain energy from respiration for maintenance purposes but forces them to pro-
duce an anaerobic product with high yields. The limitation can be somewhere in the
metabolism. Animal cells produce lactate under glucose in excess. Excretion of citric
acid in production processes is enforced by limiting the related step in the tricarbonic
acid cycle by deficiency of trace elements. This can potentially be reached by genetic
engineering or by medium depletion of some trace elements. In other cases, substrate
in excess is used by the cells to form intracellular storage compounds. These can be
starch, PHB (polyhydroxybutyrate) or oils. To avoid changes in osmotic pressure and
high cell volume such compounds are usually formed as water insoluble granules
146 | 6 Not always so simple – the batch process reconsidered

O2
Citrate
Pasteur
Effect
Crabtree
Effect
Glucose CO2
TCC
H2O

N,P,S Ethanol

Nutrient
Limitaon Intracellular
Limitaon
PHB
Biomass

Fig. 6.10: The roughly simplified metabolic structure of aerobic microorganisms showing potential
limiting steps and examples for related product formation.

Fig. 6.11: Transmission electron microscope image of

the bacterium Ralstonia eutropha producing polyhy-
droxybutyrate (PHB), which is accumulating within the
cells as PHB granules (© J. Yu).

as depicted in Figure 6.11. Such inclusion bodies can make up more than 50% of dry
cell mass. This can be enforced e.g., by introducing another limiting step again via
the medium such as nitrogen depletion. Thus, growth is limited, leading to substrate
excess.
Now we look at simulation of batch processes under constantly high substrate
concentrations and product excretions. For these types of products the substrate up-
take is constantly in the saturation range. Furthermore, the specific substrate uptake
rate, the respiration rate, the specific growth rate, and the product formation rate are
stoichiometrically coupled by energy, redox, and carbon balance, and are nearly con-
stant during the batch process. Typical cultivation curves are shown in Figure 6.12.
The simulations are carried out under the assumption of substrate uptake limi-
tation (rS,max = 0.5 g ⋅ g ⋅ h−1 , r m = 0.0). The product needs only a little extra en-
ergy for its synthesis, e.g., a polysaccharide. As a first trial the specific product forma-
tion rate rP is related to the substrate uptake rate rS with a constant yield coefficient
yP,S = rP /rS . The residual glucose is processed as usual via respiration and growth
 6.7 Primary metabolites – products directly taken from central metabolism | 147

100

90
CX
80 Cs

70 Cp

60
CX, Cs, Cp [g/L]

yP,S
50

40

30
yP,S
20

10 yP,S
0

-10
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

(a) Culvaon me [h]

0.5 rX
rs

0.4 rp
yP,S
Specific rates rX , rs , rp

0.3

0.2 yP,S

0.1

yP,S
0.0

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

(b) Culvaon me [h]

Fig. 6.12: Cultivation curves (a) and the related specific rates (b) for an aerobic batch process, where
the product needs the same amount of energy as the biomass.

with a physiological yield coefficient y X,S,physiol = 0.5 g ⋅ g−1 . The specific growth rate
r X is lower for higher specific product formation rates due to constraint of the carbon
balance. Consequently, the apparent growth yield y X,S,apparent = r X /rS is lower and
depends on the product yield. Note that in such a batch process it is not possible to
distinguish between different limiting steps unless an overshoot metabolism is active.
148 | 6 Not always so simple – the batch process reconsidered

6.8 Serving mankind with basic needs – products needed in large

amounts

Next to fuels the production of processed food or feed is the most important process in
our society with respect to carbon flux. Since the invention of controlled usage of fire
people discovered many methods of food treatment for preservation and for increasing
nutritional value. Among these operations fermentation of vegetables, meats or milk
is an important factor in many cultures to broaden the food basis. Yoghurt is the most
important example among dairy products. Lactose is degraded to lactate by homo- or
heterofermentative bacteria; in the beginning fermentation was started by wild bacte-
ria. Thus, digestibility is enabled or at least supported while decreasing pH increases
stability against contaminations. Such processes can be regarded as batch processes,
where no fresh substrate or microorganisms are added during maturation.
Remember that also today some cultivation processes like yeast production are
performed at low pH values to prevent growth of rival microorganisms. Acidification as
an important aspect of food stabilization has other side effects. As milk sours, proteins
are subject to coagulation and the milk breaks down into curds and whey. The curd
consists mainly of casein. The whey contains lactose, minerals, vitamins and other
small proteins. Separation of the solid and the fluid phase is the first step in cheese
making. This involves pressing through a sieve or a cloth. Humans have domesticated
milk-producing animals for ten thousand years and made fermented dairy products
from them. In many archeological museums in the Neolithic section one discovers
clay pots with holes in the bottom probably meant for cheese making. The first cheeses
where not further fermented and are still today in trade as white cheese, curd cheese
or cottage cheese.
Presumably inspired by carrying fermented milk in vessels made of beef stom-
ach, mankind developed the next step, which is interestingly an enzymatic one. Ren-
net from stomachs of young ruminants (cattle, sheep, goats and others) can acceler-
ate the separation process. This was further developed during antiquity to a standard
procedure. Since the nineteenth century it has been known that the active agent is an
enzyme. κ-casein consists of a hydrophobic and a hydrophilic part, stabilizing casein
micelles in the milk. The proteolytic enzyme chymosin (323 amino acids, 36 kDa) in the
rennet breaks κ-casein at a specific point into the phosphate and calcium rich para-
κ-caseinate, which is the major component of cheese curd, and glycomacropeptide,
finally found in the whey. During the cheese making process chymosin is added to the
milk in cheese vats (Figure 6.13) when enough lactic acid is formed during ripening.
This clotting (coagulation) is the central step in curd making for hard cheese.
Other bacteria or spores of moulds are added for better taste, smell or color. Then
draining is again the next step (Figure 6.14). After aging for months or even years and
preserving the surface by a rind, grinding and salting, the cheese is ready to be mar-
keted.
 6.8 Serving mankind with basic needs – products needed in large amounts | 149

Fig. 6.13: Adding of chymosin to fermented milk

(© Schönegger Käsealm).

Fig. 6.14: Cheese harp cutting the thickened (dehydrated) curd to ease water drainage in order to get
a more solid cheese (© Schönegger Käsealm).

More than 30% of milk worldwide is used for cheese making in this traditional but
nevertheless highly interesting bioprocess consisting of up to three bioreaction steps.
Numerous different cheese products are on the market produced in a nearly equal
number of process variations, a situation similar to the one in wine making or beer
brewing. This reflects a complex interplay between local traditions, technical possibil-
ities, consumer’s expectations, and societal regulations. Small and domestic cheese
makers or cheese makers in rural regions approach the same procedure as described
above, starting with wild microbial strains for acidification and using rennet from
calves. Larger commercial factories employ starter cultures and use pasteurized milk
to avoid spreading of Mycobacterium tuberculosis. Pasteurization is obligatory in many
150 | 6 Not always so simple – the batch process reconsidered

countries and raw milk cheese allowed only under specific precautions. An ethical is-
sue is slaughtering a huge amount of calves besides their limited availability. In some
religions and for vegetarians animal rennet is not acceptable so rennet substitutes
are used. Recombinant chymosin is produced extracellularly on the basis of bovine
prochymosine DNA mainly by the filamentous fungi Aspergillus niger (GRAS) and the
yeast Klyveromyces lactis (GRAS) but also E. coli. For better secretion the protein can
be changed, or in the case of Aspergillus, be coupled to glycoamylase excretion. These
products usually contain only one type of chymosine, while the natural product con-
tains three main types. As chymosine is regarded as auxiliary production material and
not as food supplement it does not have to be declared. Nevertheless, many people
do not accept this as GM (genetic modification) free and ask for rennet substitutes
based on plant or nongenetically modified microorganism. All the operations men-
tioned above have of course an impact on taste and smell and are therefore energeti-
cally debated among cheese lovers.
Besides lactate as the final product of anaerobic fermentation, other organic acids
(carbonic acids) are employed as acidulant, chelating agents, or for flavoring in food
products. Vinegar, with the main compound acetate, is next to ethanol one of the el-
dest biotechnological products in human history. The background is that (in earlier
times airborne) microorganisms grow on alcohol containing fluids (beer, wine) aero-
bically under conversion of ethanol to acetate. Nowadays industrial aerobic submers
cultivation (Acetobacter, Gluconobacter) is the standard process. The ‘substrate’ (fruit
juice, wine, spirit) is of interest insofar as it is decisive for the taste of the vinegar.
The process is not fermentation in narrow sense but a partial oxidation. Under low
ethanol concentrations Acetobacter can gain its metabolic energy by complete oxida-
tion of ethanol. So, this process can be understood as overshoot metabolism, where
under substrate saturation it is more effective for the cells to bypass complete aerobic
respiration.
Citric acid, a dicarboxylic acid, reaches a worldwide production volume of nearly
two million tons and is therefore the most important biotechnological product with
respect to market volume. It is not necessary to mention that such amounts can no
longer be made available by citrus fruits. Since 1919 and industrially since the 1930s it
has mostly been produced in submers cultivation by Aspergillus niger. Citric acid is a
central intermediate in the tricarbonic acid cycle (TCC), meaning most organisms ex-
hibit a large intracellular synthesis rate. The biologically interesting question is why
the cells excrete it. The natural purpose is to act as chelating agent for iron ions to in-
crease availability in the natural habitat. Remember that for exactly the same reason
it is part of some growth media. This happens only in trace amounts. Consequently,
an additional bottleneck has been introduced artificially. The overproduction of citric
acid requires a unique combination of unusual nutritional conditions including ex-
cess of carbon source and suboptimal concentrations of certain trace metals and phos-
phate. The combination of high concentrations of glucose and ammonium represses
the synthesis of α-ketoglutarate dehydrogenase, therefore inhibiting the catabolism
 6.8 Serving mankind with basic needs – products needed in large amounts | 151

of citric acid in the TCC and favoring its overproduction. Low manganese levels af-
fect the cell wall to enable a large flux of citric acid into the medium. Other divalent
cations influence the production as well. Furthermore, the inhibition exerted by citric
acid on phosphofructokinase, a positive end effector, has to be suppressed. Citric acid
production is therefore an example par excellence of how much control on microbial
metabolism can be exercised by the medium, here particularly by trace elements. As
no recombinant genes are involved, producing strains have been developed by induc-
ing mutations in parental strains using mutagens. Among physical mutagens, g-ra-
diation is applied. The low pH value during the production phase (pH ≤ 2) reduces
the risk of contamination by other microorganisms and inhibits the production of un-
wanted organic acids (gluconic and oxalic acids).
The theoretical yield is 112 g of anhydrous citric acid per 100 g of sucrose due to
CO2 fixation. Care has to be taken during gassing to avoid stripping of CO2 . However,
in practice, the yield of citric acid often does not exceed 70% of the theoretical yield.
The final yield after approximately one week is up to 140 g/l of citric acid and 10–
15 g/l of dry biomass. Besides for food, other applications (20% of total production)
are in the pharmaceutical industry as antioxidants to preserve vitamins, effervescent
tablets, pH correctors, and blood preservatives, or in the form of iron citrate as food
supplements, ointments, and cosmetic preparations. In the chemical industry it is em-
ployed as a foaming agent for the softening and treatment of textiles. In metallurgy,
certain metals are utilized in the form of citrate. Citric acid is also used in household
detergents as a cobuilder with zeoliths. This is supported by public regulations in or-
der to act as a phosphate substitute, avoiding eutrophication of the water body. Other
biotechnologically produced carbonic acids are tartaric, itaconic, DL-malic, fumaric,
succinic, and glucuronic acids.
In industry the ‘molecules’ to be produced are divided into ‘performance mole-
cules’ and ‘platform molecules’. The first group are typical biotechnological products
like enzymes or vitamins (vitamin B or C). These are sold not only in the health or
food sector for their biological activity, but also for technical applications. Enzymes
for washing powder have already been mentioned. Another interesting case for tech-
nical use is hydrophobin, a small fungal protein able to organize itself in monolayers
on surfaces, making them hydrophobic. During a walk in the forest it can be observed
that water easily drips off mushrooms. In particular, the spores are very hydrophobic.
The second class, the platform molecules, are produced in larger amounts for further
processing e.g., polymerization. This group includes, besides some of the carbonic
acids mentioned above, 1,4-butandiol (BDO), ‘bio’-methyl metacrylate (MMA for mak-
ing of PMMA), 1,3-propandiol, 3-hydroxy-propionic acid, or 1-octanol (fatty alcohol
for perfumes and flavorings). All these compounds can be either produced chemically
or biochemically. So, a careful debate, ‘bio versus petro’ has to be held. Decisive ar-
guments are carbon and energy efficiency and arguments regarding the raw material
(e.g., agricultural wastes) and production wastes. On the process level of course pro-
cess intensity (productivity and final product titer) play an important role. In down-
152 | 6 Not always so simple – the batch process reconsidered

OH
OH
HO O O
O O O
OH
OH
O
Acetate O O O
OH
Pyruvate HO O
O OH OH
O O
O HO HO
O
HO
O
HO

Fig. 6.15: Chemical structure of some polysaccharides with microbial and/or biotechnological origin.

stream processing separation is an issue, as 1-octanol e.g., is insoluble in water at

concentrations above 0.5 g/L and can therefore comparatively easily be gained from a
separate phase swimming on the medium. Of course costs are in the focus, where the
selling price is measured as $/MMBtu (MM million, BTU British thermal unit, cost per
heat of combustion). While ethanol is in the range of 28, 1-octanol is at 43. A high value
is a good argument for ‘bio’, justifying the greater efforts. Only a couple of molecules
(about ten according to official organizations) have currently won the race towards
‘bio’ or are candidates with good prospects.
Polysaccharides (PS) are integral parts of all cell walls (except animals) and fulfill
functions such as storage compounds, mediators of spatial structure, and protective
layers (mucus). PS are polymers of monosaccharides with the same or with different
repeating units. They can either be linear or branched, or can carry additional resid-
uals. Some chemical structures are shown in Figure 6.15. In aqueous solution they
exhibit a linear or coagulated structure or can even form helices. Their mutual inter-
action via hydrogen bonds, water binding capacity via dissociated residuals, and their
high molecular weight convey a high viscosity. For the same reasons strong physical
binding to proteins can occur. Chemical synthesis is possible, but organisms can do
it better or at least cheaper with respect to the monomer, the glycosidic linkage (α, β,
etc.), and sequence.
As manifold as their role in nature is their application as products. Quite many
PS are gained from plants. These include starch, paramylon (e.g., unicellular alga
Euglena), pectin, gums, or cellulose. The macroalgae derived polysaccharides agar,
agarose, alginate and carrageenan are used as thickeners, gelling agents (agarose gel),
or as stabilizers in food. In particular, algae cell walls differ from those of terrestrial
plants as they contain uncommon polyuronides and PS that may be methylated, acety-
lated, pyruvylated or sulfated. As these are natural compounds, there are unconfirmed
reports that e.g., carrageenan exhibits adverse health effects. The latter one is an ex-
 6.9 Conclusions for batch processes – dealing with pros and cons | 153

ample of a sulfated PS. The unicellular red alga Porphyridium cruentum produces a sul-
fated galactan exopolysaccharide with a sulfur ratio of up to 10%, which can replace
carrageenans in many applications. Also sulfated is echinacin from the purple cone
flower used as an anti-inflammatory agent and immunostimulant, and was already
applied by the indigenous peoples of North America. Some, like chitosan (shrimp
shells) and hyaluronic acid (cockscomb), used in cosmetics to rejuvenate the skin,
come from animals. Here an ethical issue applies and some efforts are made to replace
extraction from animal waste with microbial production. Already bacterially produced
in batch (due to high viscosity) is xanthan (Xanthomonas campestris) as a food thick-
ener. Harvesting is carried out by filtration subsequent to precipitation with organic
solvents. Xanthan is not digested in the intestinal tract and is sometimes thought to
cause allergic reactions. Dextran has medical use as anticoagulant and in affinity chro-
matography. Fungi derived products are scleroglucan (moisturizing, sensory charac-
teristics in personal care products) and schizophyllan, which is used in tertiary oil
recovery, where it ousts the oil from deposits. Even cellulose, the most abundant PS
in nature, is not only derived from plants but can be produced by bacteria to form thin
layers for wound dressing. Here the unique selling point is spatial structure, which
makes the difference between a low cost and a high value product.
Less emotional aspects compared to food safety apply in the discussion of bio-
plastics. At the center of endeavors are more aspects of practical applicability. The
term ‘bio’ means either derived from renewable resources, biodegradable, harmless
in direct contact with living organisms, or in a narrower sense, produced via biopro-
cesses. The discussion on biodegradability has recently become more germane, as the
tremendous problem of plastics in world oceans became more and more obvious. Be-
tween 5 and 10% of fossil fuels are deployed to produce plastic materials. This is a
great motivation to seek alternative and renewable resources. Favored raw materials
are wastes from agriculture containing lignocellulose or starch. The role of bioprocess
engineering is here finding new possible routes of conversion from wastes to valuable
products with the positive qualities mentioned above. Polyhydroxybutyrate (PHB) was
one of the first commonly accepted products with some applications. Recently, poly-
lactic acid has achieved wide application. It is produced from lactate via the cyclic
diester lactide and subsequent polymerization and consequently called polylactide.
It is questionable whether polylactide is really biodegradable under environmental
conditions.

6.9 Conclusions for batch processes – dealing with pros and cons

The batch process is still the most favored process due to its technical simplicity. In
this chapter some aspects that make it more complicated have been mentioned. Exam-
ples are undetected limitations, diauxic growth, or product inhibition. Other possible
problems are collected in Table 6.3 and discussed as pros and cons. Some of the prob-
154 | 6 Not always so simple – the batch process reconsidered

Tab. 6.3: Advantages and disadvantages of batch processes.

Physiological aspects Technical aspects

Growth rate close to μmax +++ Simple and robust +++
Permanent adaptation −− Average productivity ++
Substrate inhibition −− Average yield +
Product inhibition +/− Only minor sterilization problems ++
Undesired coproducts − Low technical maintenance +
Early uptake of cosubstrates − Batch harvesting +/−
High viscosity ++

lems can be handled with careful observation and adjustment on the medium and the
cell level. Other problems can be solved only by different process strategies as given
in the next chapters.
Medium conditions change permanently during the batch process. It starts from
inoculation where the conditions in the preculture cannot be simply kept identical to
the conditions in the fresh medium. Nevertheless, they should be as close as possi-
ble. With decreasing substrate concentration and increasing product concentration a
permanent adaptation of the cells is required at the cost of yield and volumetric pro-
ductivity, which is an intrinsic problem of the batch concept. As a batch culture is
usually started with low biomass concentrations the volumetric productivity is low in
the first half of a batch. This problem can be tackled in principle by a higher volume
ratio of the inoculum. Instead of using the harvest from one scale as inoculum for the
next larger scale, there is another idea. We harvest only a small portion e.g., 1/3 of the
volume and fill up with a respective volume of fresh medium in the same reactor. This
approach is called ‘repeated batch’ and partially overcomes the mentioned problems
of substrate inhibition, low biomass at the beginning of the process, and adaptation.
Cosubstrates and byproducts can cause problems. Initial ammonia concentra-
tions in stoichiometric amounts may cause inhibition. Uptake of ammonia leaves a
proton back in the medium leading to a decrease in pH value. The same holds for pro-
duction of acids as byproducts (see next paragraph). In cases where essential amino
acids or other cosubstrates like yeast extract are provided in the medium, the cells
can take them up faster than necessary as carbon or nitrogen sources, leading to a
depletion in a later fermentation stage. Such problems require repeated or continu-
ous dosage of the respective compounds. Besides inhibition of the main products also
other products, which are not all known, can contribute to inhibition. Experiments
have shown that intentionally added ethanol to an ongoing ethanol fermentation
leads to less inhibition than expected. The reason is that products like acetaldehyde
or acetoin are produced in smaller amounts together with the main product ethanol.
With respect to inhibition by the main product selection of robust strains showing
less effects has been done for decades. Nevertheless, it is still an issue. For economic
reasons high concentrations are mandatory. Here the batch is not that bad as it allows
 6.10 Questions and suggestions | 155

for production even at zero growth conditions or with dying cells as is the case in wine
making.
The mentioned problems may be even stronger in the case of crude substrates like
hydrolysates of wastes from the food industry or agricultural residues like straw. Mod-
ern approaches try to find a ‘robust’ cell factory. This means maintaining cell factories
in the presence of perturbations leading to less inhibition, better performance, and
optimized costs: in general better use of the cell’s inherent potential. Action can be
taken on the genetic and the metabolic level. Potentially inhibiting fractions in the
medium can be metabolized by providing the cells with appropriate metabolic path-
ways. This is also in the interest of higher yields. For the suppression of potentially
inhibiting byproducts knock-out of the related genes can be considered.
As in the example of yeast growth on glucose, many other microorganisms show a
so called overshoot metabolism under high substrate concentrations. E. coli produces
acetate, which also has a negative impact on growth. Mammalian cells convert a large
part of the glucose to lactate, an effect happening also in the muscles of living animals
during phases of vigorous exercise. To start with low concentrations to avoid substrate
inhibition and byproduct formation would lead to low cell and product concentration
at the harvesting time and is therefore not feasible. Finding a remedy is reliant on
keeping the substrate concentration low during the whole fermentation process. How
this can be practically achieved is the topic of the next chapter.

6.10 Questions and suggestions

1. In the batch simulation (Figure 6.12) it can be observed that a higher yP,S
leads to a higher product yield and a higher final product titer but also to
a longer duration of the cultivation. This reduces the volumetric productivity.
Imagine you could change yP,S by means of genetic engineering but at cost to
y X,S . What would be an optimum value with respect to volumetric productiv-
ity PV,P ?
2. Find ideas to improve bioethanol production and discuss them with respect
to technical and societal limitations.
156 | 6 Not always so simple – the batch process reconsidered

1. Assuming constant y X,S and constant yP,S during the whole process (no mainte-
nance, neglected k S ), carbon allocation may be simplified as r X + rP = 0.5 ⋅ rS
leading to y X,S + y X,P = 0.5. The end of the fermentation is given by complete
substrate uptake:
y X,S ⋅c S,0
ln ( c X,0 )
y X,S ⋅r S ⋅t end
c X,0 e = y X,S ⋅ cS,0 → tend = (6.11)
y X,S ⋅ rS

Volumetric productivity is now given as:

(0.5 − y X,S ) ⋅ rS ⋅ cS,0

PV,P = (6.12)
tend
The optimum is then given as:

dPV,P
=0
dy X,S
r2S ⋅ cS,0 ⋅ y X,S r2S ⋅ cS,0 ⋅ (0.5 − y X,S ) rS ⋅ cS,0 ⋅ (0.5 − y X,S )
→ y X,S,opt = − − +
y ⋅c y X,S ⋅c S,0 2 y X,S ⋅c S,0
ln ( X,Sc X,0S,0 ) ln ( ln ( c X,0 )
c X,0 )
(6.13)
r2S ⋅ cS,0 ⋅ y X,S rS ⋅ cS,0 ⋅ (0.5 − y X,S )
→ y X,S,opt ≈ − + = 0.25 (6.14)
y ⋅c y X,S ⋅c S,0
ln ( X,Sc X,0S,0 ) ln ( c X,0 )

Employing more precise numerical calculation leads to y X,S = 0.2 in this exam-
ple. However, it is bad news. There are not many degrees of freedom to shift the
optimum of productivity towards higher product yields. Optimization of the two
objectives is a compromise, a tribute to autocatalysis, where the ‘catalyst’ biomass
and the product are built up by the same educt.
2. The use of antibiotics during saccharification could reduce contamination risk.
On the other hand the coproduct DDGS could no longer be sold as feed, because
an uncontrolled antibiotic dosage for the cattle is dangerous and forbidden.
A simultaneous hydrolyzation and fermentation step could be possible, as is done
in sake production. Even a genetically modified yeast bearing amylases at the cell
wall or producing them extracellularly could be an idea. However, employment
of genetically engineered organisms is forbidden here.
To minimize energy for cooling the fermentation and heating during rectification a
thermotolerant yeast strain working at e.g., 70 °C would be very helpful to remove
the ethanol already during fermentation. Unfortunately, heat and ethanol have
the same target on the cell, being e.g., membrane fluidity. So the search for such
a yeast has not yet been successful.
7 Little by little one goes far – the fed-batch process
The drawbacks of batch cultivation could eventually be circumvented by measures on
the process level. The basic idea should be to keep the cells in a given physiological
state given as a defined working point in the kinetic space. This means constant spe-
cific rates and consequently constant medium concentrations. A sensible choice of
such a working point could be to keep rS smaller but as close as possible to rS,crit . In
any case, it has to be achieved via the respective substrate concentration being small
and constant. We already know that starting with a low substrate concentration will
help only for a few minutes. So an appropriate feeding strategy of the related medium
compound has to be calculated and implemented in the process.
The first idea is to measure cS online and implement a controller feeding fresh
substrate before cS is becomes too small or stopping feeding before it exceeds cS,crit .
This implies equipping the reactor with a substrate feeding stream qS,f [L ⋅ h−1 ] as the
controlled variable. Controlling the substrate feeding concentration cS,f could be an-
other option in this direction. For this approach a suitable sensor is necessary. In case
this is achievable, there is a good chance of getting the required process without any
overshoot metabolism. A similar idea is to measure the concentration of the unwanted
byproduct, e.g., ethanol in the case of yeast production, and adjust the feeding rate to
a value that keeps the byproduct concentration very low. Some yeast factories indeed
have an ethanol sensor in the off-gas stream similar to the ones the police use in traffic
control. This is a bit more indirect but is at least an option to prevent an uncontrolled
drift of the culture into substrate overshoot metabolism. This danger is actually given,
as under complex media conditions the kinetic parameters may not be known very
precisely.
What about the idea of giving the cells exactly what they need to maintain the
desired working point, so to speak controlling rS directly to keep it at an intended
value rS,set below rS,crit ? The total amount of fed sugar has to be balanced against the
amount taken up. As this is the product rS,set ⋅ c X the strategy requires in principle
knowing the cell dry mass concentration. As measuring is not very reliable the value
can be estimated as shown below. The mentioned ideas need additional equipment for
the reactor compared to batch processes, i.e., the possibility to feed permanently fresh
substrate. Therefore, such processes are called fed-batch processes. In this chapter we
will investigate fed-batch processes first by calculation of the related reactor equations
based on kinetics and mass balances. Then we will deal with process examples, e.g.,
the running example of yeast production, and discuss finally necessary prerequisites
as well as strengths and weaknesses of fed-batch processes.

https://doi.org/10.1515/9783110315394-007
158 | 7 Little by little one goes far – the fed-batch process

7.1 Setting up process equations – the formal deduction of a

suitable feeding strategy

The structure of a fed-batch reactor is depicted in Figure 7.1. Fresh medium with the
feed substrate concentration cS,f is fed with the volumetric flow rate qS,f [lh−1 ]. As
this changes the volume we distinguish here between the nominal reactor working
(reaction) volume VR,nom and the real time dependent liquid volume VR,liqu (t). This
has to be considered when setting up the balance equations.

qS,f (tcult)
cS,f tcult
VR,liqu(tcult)

Fig. 7.1: Reactor configuration for a fed-batch process.

As no biomass is removed from the reactor, the biomass balance is the same as for the
batch process taking the biomass as a balanced state variable:

dm X (t)
RX = = r X,set ⋅ m X (t) (7.1)
dt
Note that here the intended μset is used assuming that we finally manage to keep it
constant at the desired value.
For the substrate balance an additional transport term counting the freshly fed
substrate is necessary:

dmS
RS = = −rS,set ⋅ m X (t) + qS,f (t) ⋅ cS,f (7.2)
dt
As the active volume in the reactor changes by the medium inflow a third differential
equation for the changing reactor volume VR (t) is necessary:

dVR
= qS,f (t) (7.3)
dt
Because the volume changes, we can no longer obtain differential equations for the
concentrations simply by dividing the mass related equations through VR . Neverthe-
less, m X and mS can be substituted by m X = c X ⋅ VR and mS = cS ⋅ VR leading to:

d(c X (t) ⋅ VR,liqu (t))

= r X,set ⋅ c X (t) ⋅ VR (t) (7.4)
dt
 7.1 Setting up process equations – the formal deduction of a suitable feeding strategy | 159

and
d (cS (t) ⋅ VR,liqu (t))
= −rS,set ⋅ c X (t) ⋅ VR,liqu (t) + qS,f (t) ⋅ cS,f (7.5)
dt
The next step is applying the product rule for differentiation, yielding:
dc X (t) dVR,liqu (t)
⋅ VR,liqu (t) + ⋅ c X (t) = r X,set ⋅ c X (t) ⋅ VR,liqu (t) → (7.6)
dt dt
dcS (t) dVR,liqu (t)
⋅ VR,liqu (t) + ⋅ cS (t) = −rS,set ⋅ c X (t) ⋅ VR,liqu (t) + qS,f (t) ⋅ cS,f
dt dt
(7.7)

Now it is possible to normalize the equations to the volume and reorder, yielding:
dc X (t) qS,f (t)
= r X,set ⋅ c X (t) − ⋅ c X (t) (7.8)
dt VR,liqu (t)

Note that dVR /dt has been substituted by qS,f (t) according to Equ 7.3. The second term
on the right hand side represents the dilution of the biomass already present by freshly
inflowing medium. Similarly, we can write for the substrate balance equation:
dcS (t) qS,f (t)
= −rS,set ⋅ c X (t) + ⋅ (cS,f − cS (t)) (7.9)
dt VR,liqu (t)

The balance equations themselves do not tell us something substantially new, but they
can be employed to get a clearer view of what has to happen in the reactor. We remem-
ber that the initial goal of the fed-batch process was to keep cS small and constant. So
we can use these equations to calculate qS,f so that dcS /dt = 0:
! qS,f (t)
0 = −rS,set ⋅ c X (t) + ⋅ (cS,f − cS (t)) → (7.10)
VR,liqu (t)
rS,set ⋅ c X (t) ⋅ VR,liqu (t)
qS,f (t) = (7.11)
cS,f − cS (t)
Because μ is constant, m X (t) = c X (t) ⋅ VR,liqu (t) is known as exponentially increasing.
By substitution rS = r X /y X,S , and finally the desired formula for the feeding rate is
obtained:

μset ⋅ m X,0
q S,f (t) = eμset ⋅t (7.12)
y X,S (c S,f − c S (t))

This looks reasonable, as an exponentially growing culture has to be fed exponentially

to supply each cell with the same amount of sugar during the whole process duration.
To set up the formula in a process computer the yield and the initial biomass has to be
known. The initial biomass concentration and the initial volume are design parame-
ters to tune the process versus additional goals.
To test this exponential feeding strategy, it is advisable to first make a virtual experi-
ment with the standard physiological model; the result is shown in Figure 7.2.
160 | 7 Little by little one goes far – the fed-batch process

50 16

Glucose cs [g/L]

Inflow q [L/h]
14
40

12

30
10

20 8
Biomass cX [g/L]

Volume VR [L]
6
10

0
2

-10 0
0 2 4 6 8 10 12 14 16 18 20 22 24

Culvaon Time [h]

Fig. 7.2: Simulation of a simple fed-batch process with exponential feeding; the simulation param-
eters are rS,max = 2.0 [h−1 ]; kS = 1.0 [g ⋅ L−1 ]; y X,S = 0.5 [g ⋅ g−1 ]; c X,0 = 5.0 [g ⋅ L−1 ]; c S,0 = 1.0
[g ⋅ L −1 ]; V R,0 = 1.5 [L]; c S,f = 100.0 [g ⋅ L−1 ]; q S,f = 0.0225 [L ⋅ h−1 ].

As expected, the feeding rate starts from a constant value and increases exponentially.
Consequently, the filling volume of the reactor increases exponentially as well. After a
short time the initial substrate concentration stabilizes on a constant low value, which
was the intention of this feeding strategy. The biomass concentration obviously does
not follow in an exponential way. This is explained by a look at the biomass balance
equation (8.14). The first term r X ⋅ c X is the growth term, which would lead to exponen-
tial growth, and the second term −qS,f /VR ⋅ c X describes the dilution of the biomass
present by freshly fed medium. This term becomes more and more dominant during
the cultivation leading to a weaker and weaker increase of the biomass curve.
To evaluate data from a fed-batch process in order to calculate r X and rS , the total
biomass and substrate in the reactor has to be entered into the appropriate formula:
m X (t) = c X (t)⋅VR (t) and mS (t) = cS (t)⋅VR (t), respectively. The productivity is meant to
give a measure for the cost efficiency of the process e.g., with respect to the investment
into the reactor. Therefore, for the fed-batch process it is usually related to the nominal
working volume:
m X,end − m X,0 c X,end ⋅ VR,end − c X,0 ⋅ VR,0
PV,X = = (7.13)
VR,nom VR,nom

The nominal working volume VR,nom [L] of the reactor can be set equal to the final
filling volume VR,end as it is done in batch cultivations as well. This definition has
to be evaluated for the whole process or for time intervals under consideration. It is
 7.2 Running example yeast production – an industrial feeding strategy | 161

spontaneously clear that in the beginning of the fermentation this value is extremely
low due to low biomass concentration and low filling volume. Nevertheless, it is a
standard strategy for many products.

7.2 Running example yeast production – an industrial feeding

strategy

One example for a fed-batch process is the production of baker’s yeast, where over-
shoot metabolism has to be avoided. This would occur typically at specific growth
rates μ > 0.2 to 0.23 h−1 . According to the specific demands and abilities of yeasts
some specific modifications can be found in industrial production processes. Process
control policy needs careful integration of biological and technical knowledge to pro-
cess control design.

7.2.1 Deviations from exponential – controlling physiology and obeying technical

constraints

Looking at an industrial feeding strategy as plotted in Figure 7.3 several characteristic

features can be observed.
The first thing to mention is that an exponential feeding strategy has not been
applied but a piecewise linear dosing curve. This has to be understood against the

2400
70
qS,f [L/h]

60
2200
50
VR [L]
cP [g/L];

40
2000

30

20 1800
cx [g/L];

10

0 1600

-10
0 2 4 6 8 10 12 14 16

Culvaon me [h]

Fig. 7.3: Industrial feeding strategy for production of baker’s yeast; the data refer to a reactor with
2500 L working volume and a glucose concentration of 400 g/L in the feed.
162 | 7 Little by little one goes far – the fed-batch process

background of technical constraints, biological needs and last but not least desired
product qualities. The first linearly increasing feeding phase is only an approximation
of an exponential function. This is regarded as easier to realize under harsh process
conditions. The specific growth rate is indeed not constant but increases in the begin-
ning and declines at the end of the phase reaching its maximum (< μcrit ). A technical
point is that the range between the lowest and the highest value of this phase (9 to 65)
can be implemented with one pump only.
The second phase shows a constant feeding rate. This leads to a decrease of the
specific growth rate. The reasons why constant feeding is nevertheless applied lies in
a limitation of the reactor. In Chapter 6 (gas transfer) it was already noted that some-
times oxygen transport capacity is the bottleneck of productivity. This is indeed the
case here. Why substrate feeding has something to do with oxygen supply becomes
clear when recalling intracellular stoichiometry:

RS = qS,f ⋅ cS,f = rS ⋅ c X ∼ r X ⋅ c X ∼ rO2 ⋅ c X = RO2 = OUR = OTR (7.14)

A maximum oxygen transfer rate of the reactor consequently means a maximum

amount of substrate supply, which the cells can process via growth and respiration,
and thus a constant feeding rate. Note that constant feeding leads to a linear increase
of biomass concentration as the biomass integrates the feeding, so to speak. This
means in effect that the more biomass is present in the reactor the lower is the possi-
ble specific growth rate. This makes a reduction of μ unavoidable as soon the biomass
reaches a given threshold, otherwise the Pasteur effect would occur.
The third phase is characterized by a linearly decreasing feeding rate leading to
even lower specific growth rates. This stronger substrate limitation is brought about
intentionally to force the cells to take up less preferred sugars and eventually produce
traces of ethanol. Even small amounts of carbon sources in the molasses have to be
used for higher yields to make the production process more cost effective. But the di-
minishment of the growth rate has another deeper meaning.
Not all yeast is equal. The presence of buds and mother cells is a characteristic fea-
ture of unequally budding yeast cells. The fraction of budding cells (FBC), also called
budding index (BI), is actually a point of product quality for the bakery. Matured cells
are more stable in the dough and show a higher fermentation activity than the buds.
One target of the production process is to deliver a yeast product with a high ratio of
matured cells. While high growth rates contribute to high productivities and lead to
many buds, low growth rates favor matured cells. This is realized by a low growth rate
towards the end of the cultivation. FBC shows values in the relevant range for produc-
tion of 0.3 for μ < 0.1 [h−1 ] and 0.7 for μ > 0.2 [h−1 ].
The background is the cell cycle of budding yeasts as schematically drawn in Fig-
ure 7.4.
The haploid cell starts its life in the unbudded daughter phase by increasing cell
size. This takes the time T d depending on the growth conditions. Then the first check-
point ‘start’ is reached. Further action is not taken until the cell has measured its own
 7.2 Running example yeast production – an industrial feeding strategy | 163

35
pH

30
T°

25

20
cH3PO4

15

10
cNH3

0 2 4 6 8 10 12 14 16

Process me tp [h]

Fig. 7.4: Additional parameters of the industrial feeding strategy to be used controlling yeast
growth.

size and the medium conditions. Only if both are satisfactory does the cell start the
transition into the S phase. Here nucleic acids are produced, the chromosomes divide
and the cell becomes diploid. The time Tp depends on the medium but also takes some
time for the chromosomes to double. At the next checkpoint correct doubling of the
chromosomes is checked, then the cell decides to start budding. In the budded phase
the bud emerges and increases its size firstly without having its own cell nucleus. The
next checkpoint gives the green light for mitosis and the start of the meta-ana transi-
tion. This takes a defined time independent from the medium conditions and needs
intracellular storage compounds like glycogen, the major intracellular storage carbo-
hydrate in yeasts. After the bud has gained its own nucleus, it is separated from the
mother cell, leaving it in the unbudded parent phase.
Consequently, low sugar concentrations can keep the cells resting in the G1 phase
and the fraction of budding cells decreases. High temperatures also stop further cell
development at the ‘start’ checkpoint and lead to accumulation of the disaccharide
trehalose as a stress response acting as heat protection to maintain the structural in-
tegrity of the cytoplasm. In process engineering terms this is important for some steps
in downstream processing like drying and freezing. Feeding of N and P (Figure 7.5) is
a further measure to support manipulation of the fraction of budding cells. Both nu-
trient components are supplied in fed-batch, here with constant rate. This prevents
initial inhibition or precipitation. Stopping feeding at a determined point controls the
protein concentration of the final product. This can vary between 40% protein at a
C : N ratio [g ⋅ g−1 ] of 20 : 1 and 50% for C : N 10 : 1. The segregation of the cells into
the different phases is also modified. Stopping P feeding prevents nucleic acid syn-
164 | 7 Little by little one goes far – the fed-batch process

Check G1 Tc
p(np) p(j) p(i) d(nd) d(1)
d(i)
!
Tp Td
Unbudded parent phase Unbudded daughter phase
(S1-phase) (G1-phase)

!
b(1) b(k) Check G2-M
b(nb)

Budded phase Tb
(G2,M-phase)

Fig. 7.5: Cell cycle model of yeast; the checkpoints are used in production to keep the cells in de-
fined morphological states. The intervals give time points for simulation as shown in the supple-
mentary material.

thesis and keeps the cells in the S phase. Concluding this process strategy determines
not only the overall specific growth rate via substrate uptake kinetics, but also a mea-
sure used to influence cell morphology and product quality via medium concentration
measured by cell sensors.
This cell cycle is strongly conserved in eukaryotes and is therefore regarded as a
model for animal cells and even humans. Transition into a new budding phase de-
pends also on the number of buds a cell has already produced, visible by the budding
scars. This can be taken as a first sign of individual aging in organisms. In this exam-
ple, the integration of biological knowledge into process design is obvious. This has
direct impact on product qualities important for bringing the product into the mar-
ket depending on societal circumstances. A yeast for baguette baking is different from
one used for bread in Caucasus e.g., with respect to protein and glycogen content. The
necessary time of dough rise can differ between 20 min and a few days with different
demands for yeast stability. Many ideas for genetic engineering can be thought of to
make the process less complicated. One of them could be to induce ethanol formation
and the respective CO2 production only in the dough when triggered by a compound
therein. Other things might be manipulating the different checkpoints. However, regu-
lation in many countries prohibits the employment of genetically manipulated yeasts.
On the other hand, production of ‘bio yeast’ is an option. To get the ‘bio’ label strict
demands are set on the molasses: even ammonia from a chemical plant is forbidden
and natural protein hydrolysates are used as N source in some factories.
 7.2 Running example yeast production – an industrial feeding strategy | 165

7.2.2 The first step of downstream processing – relationship between cell

separation and cell physiology

Harvesting is from the view of mechanical process engineering a solid/liquid separa-

tion task to be accomplished by different centrifugation and filtration steps. To under-
stand what is specific for bioprocess engineering the basic formulas are compiled:
Stokes law gives the relation between the force of viscosity Fdrag [N] on a small
sphere with (hydrodynamic) radius rpart [m ⋅ s−1 ] moving through a fluid of dynamic
viscosity ηfluid [Pa ⋅ s] with the velocity vpart [m ⋅ s−1 ]:

Fdrag = r ⋅ π ⋅ rpart ⋅ ηfluid ⋅ vpart (7.15)

Furthermore, gravity and buoyancy act on a particle according to the density ρ

[kg ⋅ L−1 ] with the resulting force:
4
F g = (ρ part − ρ fluid ) ⋅ g ⋅ ⋅ r ⋅ π ⋅ r3part (7.16)
3
The equilibrium of forces finally leads to the Navier–Stokes equation for sedimenta-
tion velocity vS [m ⋅ s−1 ]:

2 (ρ part − ρ fluid )
vS = ⋅ g ⋅ r2part (7.17)
9 ηfluid
Now we can discuss this equation with respect to cells. The first point is the radius
of the particle. For bacteria we can assume here 1 µm and for eukaryotic cells e.g.,
5 µm, so low sedimentation velocities are expected compared to other technical parti-
cles. For bacteria sedimentation velocity can even be a factor 25 lower as the velocity
increases with the square of the radius. The second point to be considered is cell den-
sity. This can be roughly estimated to be 1.1 [kg ⋅ L−1 ]. This value is much lower than
values e.g., for mineral particles. Cell density may change with macromolecular cell
composition. A high oil content can reduce the difference in density between cell and
medium nearly to zero. As an example the sedimentation velocity of a yeast cell yields
with rCell = 5 μm, ∆ρ Cell = 0.1 [kg ⋅ L−1 ], ηH2O = 1 [mPa ⋅ s] and g = 9.81 [m ⋅ s−2 ]
the value of v s,yeast = 5.45 ⋅ 10−6 m ⋅ s−1 . So we can wait for hours until a sedimen-
tation can be observed in a measuring cylinder. As a third point one has to be aware
that media are not water but can have much higher viscosities up to a factor 1000 if
polysaccharides are present. All three points make it obvious that specific measures
have to be taken. In a centrifugal field sedimentation is faster by the factor z, the rel-
ative centrifugal force. This factor can reach for technical centrifuges values of up to
15,000. Even then centrifuges with very short sedimentation lengths are necessary.
For yeasts disk stack centrifuges – known from milk processing – are usually chosen,
producing the yeast cream as indicated in the introduction Figure 1.8.
For bacteria even this is difficult, so structural changes have to be envisaged. Floc-
culation is one commonly applied means, increasing the hydrodynamic diameter from
166 | 7 Little by little one goes far – the fed-batch process

cell to aggregate level. This can be achieved for bacteria with flocculants i.e., highly
charged macromolecular substances. However, it is not comfortable to have another
substance in the product, which has to be removed later. As bioengineers we start
looking for a solution in biology and find that many microorganisms are able to show
flocculation naturally. In many industrial applications like production of bioethanol
Saccharomyces show flocculation at desired points during the process, which is an
important characteristic of a good production strain. Yeast flocculation depends on
the expression of specific flocculation genes (FLO1 to FLO11). Expression of the floc-
culation genes is influenced by the nutritional status of the yeast cells as well as other
environmental factors (e.g., T, pH, pO2 , Zn) and can therefore be controlled by process
design facilitating later separation. Proteins need values of 200,000 g for centrifugal
separation, which is only possible in lab scale. Flocculation e.g., by change of pH value
or increasing salt concentration is possible (reducing surface charge and zeta poten-
tial) and applied for large scale separation of proteins (‘salting out’).
For production of yeast cubes a second concentration step is necessary. Vacuum
drum filter is the means of choice. Darcy’s law describes the area specific flow of the
filtrate through the filter cake assuming it as a porous medium:
Qfluid κ ⋅ ∆p
= (7.18)
Acake dcake ⋅ ηfluid

This is a linear law between the driving force ∆p and the volume flow Qfluid modified by
geometrical constraints. So far it is similar to other transport equations (e.g., Fourier’s
law for heat and Fick’s law for diffusion). Material specific coefficients are viscosity of
the fluid ηfluid and the intrinsic permeability κ [m2 ]. The parameter κ depends basi-
cally on the particle size and shape. In practice it cannot be calculated precisely at the
cells are not spheres, they may be deformed, or they may be biospecific, and the gus-
sets between the cells may be filled with extracellular material (EPS) impeding filtrate
flux. An experiment will give reliable values. Results are shown in Figure 7.6. Measured
is the specific cake resistance αcake = dcake ⋅ κ−1 /(mcake ⋅ A−1 −1
cake )[m ⋅ kg ].
The first thing we notice is that the cake resistance increases by a factor of ten with
increasing pressure drop by a factor of ten as well. This means that despite of Darcy’s
law filtration velocity stays constant with increasing pressure. Instead of pressing the
medium through the cell scaffold the cake is compressed. This happens not so much
by cell deformation but by narrowing the distance between the cells leaving less space
for the filtrate flux. Filter cakes showing such a behavior are called ‘compressible filter
cakes’. Furthermore, budded cells show a higher resistance than unbudded cells. This
is another example of a direct functional chain from cell morphology over feeding
strategy to behavior in downstream. So unbudded cells are preferred not only for high
fermentative activity in the dough but also better filterability.
Filter cake resistance turns out to be so high that only very thin filter cakes are
obtained. So again a structural change in the system has to be brought about. This in-
cludes precoat filtration (Figure 7.7) employing diatomaceous earth as filter aids. This
 7.2 Running example yeast production – an industrial feeding strategy | 167

unbudded yeast
budded yeast
Specific cake resistance [m/kg]

1E13

1E12

1 10
ΔP Differenal pressure [bar]

Fig. 7.6: Measurement of specific cake resistance of unbudded (straight line) and budded cells
(dashed).

Feed:
Cells and filter aid

Coang
Filter medium

Fig. 7.7: Principle of filter aid filtration with diatomaceous earth; the REM pictures (small ‘tons’)
show the fossil frustules (‘shells’) of Melosira.

material allows for better water permeation and better mechanical removal of the cake
from the filter cloth.
This approach is also borrowed from the food industry, namely fruit juice filtra-
tion. Interestingly, diatomaceous earth is a biological product formed by diatoms (mi-
croalgae) during past geological eras. Artificial particles would hardly exhibit such
microscale structures. Despite all our technical sophistication the implication of the
choice on humans should not be forgotten. The material is mined from fossils deposits.
The dust is suspected to be a cancer promotor for the workers. Furthermore, it may
contain heavy metals, which makes it necessary to bake the material, which increases
costs. Finally, the used precoat material has to be disposed partly in landfills for spe-
168 | 7 Little by little one goes far – the fed-batch process

cial wastes. This is reason enough to look for new solutions to precoat filtration, in the
best case biodegradable and biogenic particles, e.g., based on lignocellulosics.
The product from the filtration step is still a pasty and not a crumbly mass as
expected for yeast cubes. Here another difference between cells and mineral parti-
cles comes into action. By washing steps the osmotic environment of the filter cake
is changed so that the yeast takes up the residual gusset water leading to the desired
consistency of the ‘fresh yeast’ cake. Further, granular ‘active dry yeast’ (or ‘instant
yeast’) is obtained by spray drying, making the product more durable. Freeze drying
is standard for keeping in strain collections.
Filtration has also been invented by nature where some animals – filter feeders –
practice their ingestion of food by filtration with specialized morphological struc-
tures. Flamingos dig their curved beaks upside down into water to catch shrimps or
cyanobacteria. Residual water is pressed out through hairy structures called lamellae
which line the mandibles. Humpback whales feed on krill or swarms of herring. They
swim, mouths agape, swallowing thousands of fish in one gulp. Pleated grooves in
the whale’s mouth allow them to drain the water, filtering out the prey. Both animals
can build up pressures of several bars to filter their prey as effectively as possible.
As in technical filter processes a preconcentration step is helpful, therefore several
whales produce a bubble net, which confines the space for the prey (bubble net feed-
ing). The krill itself is a filter feeder. It uses a special structure on its front legs, the so
called ‘filter basket’, with which it can catch marine microalgae. Compare the section
on marine food chain in Chapter 8 on microalgae. A mussel can clean an aquarium
full of suspended microalgae within a few hours. In the realm of microorganisms sun
animalcules (heliozoans) invented sticky mucus filtration. This has not really found
a technical equivalence yet.

7.3 Production of secondary metabolites – case study of

penicillin production

Penicillin was one of the first biotechnological products for medical use and one of
the outstanding examples of biotechnology. Many books and even films have been
published about the life and work of Alexander Fleming (1881–1955). In 1928 he saw
“by pure chance” that a Staphylococcus culture on an Agar plate was contaminated
with molds. Instead of discarding the plate like probably many scientists before him,
he observed around the contaminated spot a zone free from bacteria. Further experi-
ments revealed that indeed the fungi secreted a substance that could kill gram-positive
bacteria even in 800-fold dilution. So the first thing to learn from him is to look care-
fully on the findings, even if it is not in the mainstream of the recent research. “One
sometimes finds what one is not looking for,” as Fleming said. The published paper
attracted nevertheless no further attention. In 1938, a systematic study (Ernst Chain,
 7.3 Production of secondary metabolites – case study of penicillin production | 169

Variable ß-Lactam- Thiazolidine-

side chain ring ring

H
NH S CH3

O N CH3
O
OH
O
Phenyl-Acetate L-Cysteine D-Valine

Fig. 7.8: Molecular structure of penicillin showing the eponymous cyclic amid structure and the
precursors from which the cells synthesize the completed molecule.

Howard Florey) of antibacterial activity of microbial substances led to the rediscovery

of Fleming’s findings. The background was the Second World War, where many sol-
diers did not die directly of their injuries but of secondary infections. Consequently
further development was pushed by political authorities mainly in USA and led to
technical production on surface and submers cultivation. This saved millions of lives
during the war. This is another lesson to learn: scientific interest in terms of technol-
ogy push is sometimes not enough for a useful development. In addition a market
pull is necessary in this case a high sense of affliction. The task of the bioengineer is
to build a bridge over this gap and develop new processes as effectively as possible.
Today (2015) sales of β-lactam antibiotics have reached over US$20 billion. They
get their name from the β-lactam ring structure (Figure 7.8). Product titers are with
60 g/l (=100,000 IU/mL) more than 50,000 times higher than the original strain of
Fleming, reason enough to have a look at the mechanisms applied for strain and
process improvement. As a result of the first screening steps production strains were
changed from P. notatum to P. chrysogenum. Further remarkable development was
achieved by undirected mutagenesis (e.g., X-ray, UV, yperite). These techniques can-
not lead to overexpression of certain genes (unless intracellular control loops are
deleted) and certainly won’t introduce new functional genes. Unlike primary prod-
ucts, which are limited by intracellular catabolic fluxes determined by substrate
uptake, stoichiometry, or intracellular bottlenecks, secondary products underlie a
strong intracellular regulation. The fungus produces the desired products according
to temporary physiological needs stimulated by intra- and extracellular signals. Oth-
erwise the cells save the energy they would use to do so. This could be in the case of
penicillin, a little bit speculatively, the presence of bacteria as competitors of glucose
in the environment. The success of undirected mutagenesis is based on destruction
of relevant intracellular control loops, leading to deregulation of negative feedback
inhibition. For about 20 years, further process improvement has only been reached by
employment of modern means of genetic engineering like overexpression of specific
170 | 7 Little by little one goes far – the fed-batch process

proteins, introduction of other promotors, and application of metabolic engineering

in general. Growing capacities go more and more to China and India.
Modern strains with high production rates have their limitations in ATP or NADPH
demand and the presence of specific precursors, a situation of metabolic burden not
very different from the situation in recombinant protein production. So only 10% of
the consumed carbon is targeted to penicillin, while 25% is available for growth. The
remarkably high remaining 65% is assigned to intracellular transport, product excre-
tion, and hydrolytic loss of activated intermediates. Here further progress is poten-
tially possible despite the complex regulation network involved in penicillin forma-
tion.
In parallel to strain development, success has been accomplished by rational pro-
cess design. This was guided also by insights into the biological circumstances, lead-
ing to high production rates. To understand some aspects of processes where filamen-
tous fungi are involved we have a look at their morphology. Penicillium and Aspergillus
are not unicellular organisms like bacteria. Nonetheless they rank among the microor-
ganisms. They propagate by spores and form filaments of linearly connected cells, the
hyphae (see Figure 7.9) of a few µm diameter and up to the cm range in length. Growth
occurs only at the tip of the hyphae, and cell material from the older cells is trans-
ported along the hyphae to the tip. Depending on the length of the filaments and the
growth conditions branching occurs by formation of new tips. The entirety of all hy-
phae is called the mycelium. The older cells form vacuoles of increasing volume and
finally die. As in yeast, the individual cells have different age and physiology and have
to be described in terms of a segregated model.
In nature the mycelium can cover an area of several m2 . As nutrients in the mid-
dle may be depleted, growth and fruiting body formation occurs mainly at the edge of
the circular area, a phenomenon called a fairy ring. In the bioreactor the hyphae are
exposed to shear forces and vortices bending them back to the center forming finally
a pellet of a few mm diameter. A graphical simulation is shown in Figure 7.10. This
has repercussions on the process. Nutrients and oxygen are used up in the outer layer
of the pellets, which form a diffusion resistance on top. The inner space depletes in

Degenerated cells Parally degenerated cells

Hypha: cell filament Branching Tip

Apical growth
Mycelium: Enrety of hyphae

Fig. 7.9: Formation of a hypha and branches in filamentous fungi; depicted in golden color are the
active cells, in blue water-containing vacuoles.
 7.3 Production of secondary metabolites – case study of penicillin production | 171

Spore
Hypha Branching

Bending Pellet 100 μm

Fig. 7.10: Graphical simulation of pellet formation starting from the spore and continuing over a
growing mycelium.

0.010

0.008
rp [g/gh]

0.006

0.004

0.002

0.000
0.000 0.005 0.010 0.015 0.020 0.025 0.030 0.035 0.040 0.045 0.050

 [1/h]

Fig. 7.11: Product formation kinetics, being of course strain specific.

nutrients, causing cells to die a leave a hollow in the center. This is not totally a dis-
advantage. Pellets are easier to filtrate than single cells due to their size. Some degree
of substrate limitation can also support product formation as outlined below.
The next observation is that Penicillium shows high production rates only under
low growth rates. So this is inverse compared to products from primary metabolism.
Of course enough carbon has to be present meaning that the peak of the specific pro-
duction rate is observed at low specific growth rates (Figure 7.11).
172 | 7 Little by little one goes far – the fed-batch process

80 1600 50 0.12

70 1400

Specific growth rate  [1/h]

Total glucose added ms [g]
0.10
40
60 1200
0.08
Biomass cX [g/L]

50 1000

Penicillin cp [g/L]
30
40 800 0.06

30 600 20
0.04
20 400
10
0.02
10 200

0 0 0 0.00
0 20 40 60 80 100 120
Culvaon me T [h]

Fig. 7.12: Data from a penicillin production process showing the specific growth rate in the two
phases and beside biomass and product concentration the integrated substrate feed.

Glucose in particular suppresses product formation. This gives us the opportunity to

install a two phase fed-batch process as in the case of enzyme production. During the
first phase mycelial biomass can grow up to a sufficient level. Here glucose is the best
substrate. In the second phase production is induced by reducing the concentration
of the carbon source to a level that allows only for low growth rates (Figure 7.12). Here
lactose is preferred, as it does not suppress product formation. Both sugars can be in
the initial medium, from which glucose is used up first. As penicillin is a compara-
tively cheap product, the process has to be based on cheap natural media. In the first
phase mainly molasses or other crude carbon sources can be chosen and nitrate as N
source. Feeding of corn steep liquor as additional C and N source is applied in the first
phase and continued in the second phase. We remember that corn steep liquor con-
tains mainly lactose as C source, while molasses contains glucose. Due to high prod-
uct titers nitrogen from corn steep goes to a large amount into the product. Ammonia
would suppress product formation. The second phase can be prolonged by so called
‘miniharvest’ protocols, where a part of the reactor content, e.g., 30%, is harvested
and fed batch can continue.
The next important aspect with respect to medium design is the additional feed
of precursors to support the fungi in making the product. These are namely the amino
acids L-cysteine and L-valine and for penicillin-G phenyl-acetic acid, so basically the
compounds from which the penicillin molecule is built up. This precursor leads to ‘di-
rect synthesis’, while other precursors lead to nonnatural ‘biosynthetic’ antibiotics.
 7.4 Production of recombinant proteins – the cell factory | 173

Acvated
Butyl-acetate carbon
Water Solvent
Purificaon
H2SO4
pH2

Cell mass Butyl-

acetate Extracted medium Recovered
Culvaon Cell separaon acvated carbon
Solvent-Extracon
Acetone, H2O
K-Acetate
Hot air Solvent Solvent acetone

Penicillin G
K-Salt Drying Crystal filtraon, washing

Fig. 7.13: Flow sheet of a penicillin production process from cultivation through cell separation, and
extraction to formulation.

For all side chain precursors uptake has to be ensured as well as induction of enzy-
matic activation.
Production fermenters reach volumes up to 100 m3 scale with scaling steps in the
production train of 1 : 10. A flow sheet of the whole process is depicted in Figure 7.13.
The first downstream process is cell separation by vacuum drum filtration. This is
enabled by the pellets forming a quite porous filter cake. In modern processes product
titer is high enough to enable wet extraction without a prior solid/liquid separation
step. Note that the mycelium, due to its content of residual antibiotic activity, cannot
be used as cattle feed but needs special disposal steps. Extraction occurs at low pH
values to give a high partition coefficient. Unfortunately, the product is not very stable
at this point making a fast separation of medium and solvent necessary by application
of centrifugal extractors. Finally the product runs through several purification steps.
Crystallization has high ability to exclude impurities. Finally a clean and dry penicillin
salt is obtained.

7.4 Production of recombinant proteins – the cell factory

Until now we considered whole cells and small molecules as products. While metabo-
lites have a high turnover, they occur only in small intracellular concentrations. The
cell itself is built up of macromolecules showing much lower turnover rates. Macro-
molecules carry the main functionality of a living cell. From a technical point of view
we find possible products among all groups of biological macromolecules. These
groups are basically nucleic acids, proteins, polysaccharides and lipids. Usually they
174 | 7 Little by little one goes far – the fed-batch process

are built up from a long but specific sequence out of a couple of different monomers.
This sequential order defines structure and function. In this paragraph proteins are
the focus and especially process engineering aspects of their production. The com-
plex biological machinery for replication of nucleic acids, transcription, and finally
translation into proteins belongs to the most amazing inventions of nature. In the
meantime this machinery is scientifically understood precisely enough, a success
story on its own, to allow for rational process design. Consequently, exploiting the
bioactive capabilities of proteins in food production, for pharmaceutical use, or as
catalysts in technical applications is a major field of bioprocess engineering.

7.4.1 Different ways to get hold of desired proteins – an overview of protein

expression systems

The first option is to look for ready solutions made by nature. Many organic substrates
are found in nature as water insoluble or even denatured biopolymers. Nevertheless,
they are degraded by microorganisms via hydrolytic enzymes, which are secreted
into the environment. One potent exemplar of a microorganism with this capability
is Bacillus subtilis. So quite early in the history of biotechnology it was employed for
the production of proteases (washing powder). To design a production process based
on Bacillus the same considerations with respect to the medium apply, meaning a
suitable N source not only for growth but also for the target protein. As in penicillin
production we need a biological signal forcing the organisms to start production and
secretion. It is not really astonishing that protease is produced mainly during nitrogen
limitation, while phytase is produced during phosphate limitation. As in the penicillin
case, undirected mutagenesis or directed evolution are means to destroy intracellular
feedback mechanisms between substrate availability and enzyme production. The
idea for the process is consequently again a two stage fed-batch process for getting
high cell counts in the first phase and high product titer in a substrate limited second
phase. An example is more closely described in Chapter 11 on fermentation.
It is desirable not only the potential for production of large amounts of technical
enzymes but also for its high ability to secrete them efficiently into the environment
or respectively the medium. This is an important example of integration between bi-
ology and downstream processing, as it enables easy separation of the enzymes from
the medium without cell disruption and separation of the target protein from a few
thousand other intracellular proteins. In fact, protein separation by chromatography
after preceding mechanical cell separation and disruption steps is a sophisticated,
and in the end expensive enterprise. The idea of integration is to leave such things
to the cells. Cell membranes have the most potential as selective separation devices.
However, natural extracellular expression systems, which may be bacteria, yeasts, or
fungi, are employed mainly for technical enzymes as they are not very large. Another
point of consideration is that recombinant proteins may be the target of natural pro-
 7.4 Production of recombinant proteins – the cell factory | 175

teases secreted by Bacillus, even if we do not want them to do it. Bacteria cannot pro-
duce proteins with human identical glycosylation or other forms of posttranslational
modification. Such mainly pharmaceutical proteins can be provided only by eukary-
otic organisms.
To make use of this property of eukaryotes, yeasts, fungi, and last but not least
animal cells are applied. Phototroph production in moss or microalgae is also under
development. Now we collect some points how to support the ‘cell factory’ along the
way from transcription to excretion. Firstly, we have to ‘convince the management’ to
produce the target protein and not proteins needed by the cell itself. Geneticists will
provide strains carrying the required gene in many copies if necessary. Secondly, the
promotor has to be defined to start production only at the beginning of the production
phase. From a process engineering view this is transmitting a chemical or physical
signal to the cell to activate the promotor, so something like ‘placing an order’ to the
cell factory. Some commonly applied signals are collected in Table 7.1. Some of them
are explained in more detail in the following paragraphs.
The cells are optimized for self-reproduction so forcing them to produce recom-
binant proteins will require resources of the cell factory, which are then no longer
available for the cells to proliferate. This ‘metabolic burden’ has to be compensated
as effectively as possible or tolerated only during the production phase. Gene expres-
sion should be balanced against the amount of available amino acids and cellular
ATP generation. In fact, as soon as a high cell concentration is obtained in the growth
phase, further growth is not obligatory but as much carbon as possible is allotted to
the product. The necessary amino acids are the ‘parts’ for further ‘assembling’ the
proteins. Feeding amino acids is a means to help the cell carry the metabolic burden.
The translation process is located at the ribosomes, which can therefore be called the
‘work systems’ or more precisely the ‘assembly machine’ stressing the cell factory pic-
ture. Their number depends on the type of the cell. Like in a real factory short cycle
times set under pressure of time can increase the scrap rate in the case of proteins, vis-

Tab. 7.1: Fed-batch processes, signals and induction; fill in further examples according to your own
experiences.

Organism Product Signal for Signal for Mechanism

observation induction

Bacillus Substrate Intracellular control loop

limitation
E. coli Rec. proteins pO2 peak IPTG Induction of lactose operon
Heat Heat shock proteins (chaperons)
Light Rhodopsin
Aspergillus Penicillin Target Low glucose Intracellular control loop
concentration
Animal cells pH
176 | 7 Little by little one goes far – the fed-batch process

ible as incorrect sequence like missing terminal amino acids or misfolding. Eukaryotic
proteins are subjected to further posttranslational modifications, especially glycosy-
lation, where oligosaccharides are attached to the proteins. These glycosylation pat-
terns are very specific for different groups of organisms controlling enzyme activity,
interaction with cell surfaces, or helping the organisms to distinguish between their
own and foreign proteins. Incorrect glycosylation patterns can cause malfunction of
pharmaceutical proteins in the human body. ‘Humanization’ of the proteins makes
production in mammalian cells necessary in many cases, although it is quite expen-
sive. In other cases glycosylation can be modified by subsequent enzymatic steps or
even by genetically engineering the intracellular enzyme system. Finally ‘parcel la-
bels’ have to be put on the products to allow the intracellular transport system to de-
liver the protein to the intended location, be it cytoplasm, periplasm, or extracellular
space. These targeting tags are coded in the genome as well, detected by the cell, and
cleaved after delivery; see also Chapter 2.

7.4.2 High density cultivation – a specific fed-batch process for recombinant

protein production

Recombinant proteins mostly for medical use are a big success story in biotechnol-
ogy. Some of them like insulin can be produced by bacteria. Human insulin produced
by E. coli was the first ‘golden molecule’ of the biotech industry and has been on the
market since 1982. This was only four years after the first insulin encoding gene was
transferred from mammalian to bacterial cells. Since 2000 insulin analogues with bet-
ter pharmacokinetics have been available. In addition, to help millions of diabetics
worldwide effective production processes have been developed. In contrast to extra-
cellular technical enzymes produced by Bacillus, intracellular proteins agglomerate
as ‘inclusion bodies’ (Figure 7.14.).
The process runs through three phases as shown in Figure 7.15. The first phase is a
batch phase, where the E. coli cells grow with maximum specific growth rate to mod-
erate cell densities. Concomitant acetate is produced due to the overshoot metabolism

Fig. 7.14: Inclusion bodies of GFP (green fluorescence

protein) by Escherichia coli BL21 as example of in-
tracellular accumulation of recombinant proteins. In
this study, the incorporation of bacteriotoxic peptide
(protegrin-1) into the GFP allows the expression of in-
soluble proteins in high concentration (© J. Yu).
 7.4 Production of recombinant proteins – the cell factory | 177

Batch Fed-batch
Phase 1 Phase 2
1000 30 1500 100

90  = 0.45 h -1
 = 0.11 h -1

25
100 1200
Agitaon rate nsrr [rpm] 80

70
20
10 900 60

15 50
pO2 [%]
cX [g/L]

cs [g/L]

1 600 40
10
30

0.1 300 20
5
10

0.01 0 0 0
0 5 10 15 20 25 30 35
Culvaon me T [h]

Fig. 7.15: Data from high density cultivation showing the two phases, in particular the gas phase
related values p O2 and agitator speed.

of E. coli preventing further growth to higher densities. Increasing biomass concentra-

tion leads to exponentially decreasing dissolved oxygen concentrations. A controller
prevents decrease to limiting values by increasing stirrer speed. At the end of this
phase all glucose is consumed. The idea is now to wait a short time to force the cells to
take up the acetate. These events of glucose and acetate depletion cannot be measured
directly. Now an indirect signal is used based on the substrate/respiration coupling.
The complete lack of a carbon source therefore disables oxygen uptake. This is mea-
sured as a peak in pO2 . The pO2 controller is parameterized in such a way that increas-
ing pO2 does not lead to falling stirrer speed. Sometimes different peaks are observed
indicating glucose, acetate, and glycerin depletion. Glycerin is sometimes part of the
medium coming from the deep-freeze cans in which the inoculum is transported. Af-
ter the last peak a classic fed-batch strategy is started. This first step shortens in total
the long first phase of a fed-batch process with low volumetric productivity.
During the second phase, carried out in fed-batch mode, acetate production is
suppressed by low growth rates and the culture propagates to high cell densities up
to 100 g ⋅ L−1 . Dissolved oxygen is controlled at sufficient values again by the agitator
or higher gas flow rates. As we have seen in the yeast examples oxygen transport lim-
itation is possible, especially as shear stress induced by the stirrer has to be avoided.
Gassing with oxygen enriched air is possible in so far as the high added value in con-
trast to yeast production is economically acceptable. During this phase genes for prod-
uct formation are not active.
178 | 7 Little by little one goes far – the fed-batch process

80 8

cProt [g/L]
60 6
cs [g/L]

40 4

cAce [g/L]
cX [g/L]

20 2

0 0
0 6 12 18 24 30 36
Culvaon me T [h]

Fig. 7.16: Measurements and simulations of a high density cultivation. Inductor addition triggers the
third phase, which is the production phase.

Only when high cell densities are achieved is production started by induction of the
respective genes in the third phase. Induction can be achieved by chemical signals
(IPTG) or heat induction. In this case chaperones in the form of heat-shock proteins
are produced by the cells to prevent misfolding of proteins. The gene of interest has
the same promotor as the chaperones. Overexpression of the recombinant proteins
is a high metabolic burden for the cells. All carbon and energy is allotted to the prod-
uct. Under these stress conditions again acetate is produced as shown in the measure-
ments shown in Figure 7.16. This makes it clear that simultaneous growth and produc-
tion is not a matter of choice. Low growth rates before and during the induction phase
furthermore reduces incomplete translation and incorrect folding.
From HDC we can learn two principles. The first one is to decouple physiologi-
cal states into different phases of the process. Here these states are high growth rate
but acetate formation, lower growth rate but high cell density, and product formation
without growth but good product quality. The second principle is to measure signals
of the cells and to apply physiological signals to the cells for induction.

7.4.3 Production of pharmaceutical proteins with Kamagataella pfaffii –

employment of a high performance factory

Kamagataella pfaffii, formerly known and still sometimes termed as Pichia pastoris, is
a methylotrophic yeast. It exhibits high growth rates and reaches high cell densities.
It is able to grow on a simple, inexpensive medium, which is free from all compounds
 7.4 Production of recombinant proteins – the cell factory | 179

like proteins that could potentially interfere with production or product separation.
K. pfaffii has been classified as GRAS by the American Food and Drug Administration
(FDA), where one point is that no pyrogens or lysogene viruses are known. Further-
more, it shows high levels of secretion. So, it is an attractive host for the expression of
recombinant pharmaceutical proteins. The yeast possesses two different alcohol oxi-
dases (AOX), from which AOX1 has one of the strongest natural promotors. AOX1 has
to be deleted and replaced by the recombinant gene sequence. Carbon uptake dur-
ing growth and production relies on the much weaker AOX2. Excretion is evoked by a
fusion protein e.g., the α–factor prepo signal sequence.
The process requires two substrates as visible in the process scheme in Figure 7.17.
During an initial batch phase on glycerol the cells can grow with their maximum spe-
cific growth rate (approximately 0.15 h−1 ). In the second phase, a glycerol limited fed-
batch phase is started, as in high cell density. Here the biomass reaches predefined
values of >50 g/L, while protein production is repressed. Then the production phase is
initiated. Methanol feeding leads to induction of the strong promotor and expression
of the desired recombinant protein, which is finally secreted into the medium. Another
advantage of P. pastoris, the tolerance for a high pH range, allows pH adjustment to
values favoring protein stability. For the sake of stability a continuous separation of
the proteins by filtration and cell retention can be foreseen.

FP
micro filtraon
weight
FR1
control
ultra filtraon

FR2

CS1R1 CS3R2 culvaon FMF

and
producon

glycerol feed methanol feed heang/ cooling

permeate FUF
tank waste tank

Fig. 7.17: Flow sheet of an experimental pilot plant for recombinant protein production with Kama-
gataella pfaffii.
180 | 7 Little by little one goes far – the fed-batch process

7.5 Conclusions for fed-batch processes – dealing with strengths

and weaknesses

The first idea of applying fed-batch feeding policies was to avoid initial substrate in-
hibition, overshoot metabolism, and other disadvantages of batch processes. Media
with high initial substrate concentrations can be applied without dissolution. Even
if the main substrate is not applied in a fed-batch process, it can be applied for co-
substrates. Ammonia is toxic only in comparatively low concentrations and has to be
applied in a fed-batch process to reach high biomass concentration. The same holds
for phosphate to prevent precipitation or intracellular accumulation. Furthermore, the
fed-batch process turns out to be a flexible means of controlling cell physiology (e.g.,
age, cell composition) and designing processes in different phases, especially growth
and production phase. Therefore, fed-batch processes are widely applied in different
kinds of (pharmaceutical) processes due to their flexibility and strict physiological
control. Some of the pros and cons are collected in Table 7.2.

Tab. 7.2: Pros and cons of fed-batch processes.

Physiological aspects Technical aspects

Growth rate controlled +++ Control necessary −

No substrate inhibition ++ Average productivity +
No overshoot metabolism ++ Average yield ++
Application to cosubstrates + Changing volume −
High substrate concentration ++ High technical maintenance −
Product induction +++ Batch harvesting +/−
Flexible process design ++

Technical problems include changing volume during the process and low productivity
in the initial phase. This can be avoided by a preceding batch phase or by choosing
high substrate concentrations. The ideal case would be to employ pure substrates with
only little water, which may by gaseous (methane), fluid (methanol, glycerol), or solid.
This last option is of course hardly practicable. High technical maintenance demand
and necessity of a good measurement and control concept could be a hurdle to be
managed based on existing technology and experiences. Batch-wise harvesting can
be a tribute to batch-wise downstream technologies and allows especially for charge-
wise quality control in the pharmaceutical industry.
 7.6 Questions and suggestions | 181

7.6 Questions and suggestions

1. Read more about the history of Fleming to understand the dynamics of sci-
ence and process development. Find other examples where ‘lucky findings’,
technology push, and market pull are driving forces of our endeavors.
2. Calculate r X , rS , y X,S , PV,X from the measured data in Figure 7.2: t0 = 0.0 h,
c X,0 = 5.0 g/L, cS,0 = 2.5 g/L, V0 = 1.5 L, tend = 24 h, c X,end = 45.4 g/L,
cS,end = 2.5 g/L, Vend = 15.0 L. The feeding rate qS,f is given as 0.0225 ⋅
exp(0.2 ⋅ t) l/h, cS,f = 100 g/L.
3. Sometimes constant feeding is chosen (with limitations in OTR, pumping
rate, gaseous substrate, light). How does the biomass and the specific growth
rate develop?
4. What is the theoretically maximum final biomass concentration that can be
reached in a fed batch? What happens during several cycles of miniharvest
and filling-up of the harvested volume with fresh medium?

2. Total biomass produced m X,prod = m X,end ⋅ Vend − m X,0 ⋅ V0 = 673.5 g; total sugar
consumed mS,up = ∫(qS,f ) − mS,end ⋅ Vend + m X,0 ⋅ V0 = qf,S,0 ⋅ 1/r X,set ⋅ exp(r X,set ⋅
t) = 1334 g; y X,S ≈ 0.5; PV,X = m X,prod /VR,end = 1.87 g/L/h; assuming constant
physiological conditions (cS = const.), r X = 1/∆t ⋅ ln(c X,end ⋅ Vend /c X,0 ⋅ V0 ) =
0.188/h; the specific growth rate is lower than expected, closer inspection reveals
that cS also increases slowly. The reason is the wrongly calculated initial feeding
rate, which should be 0.03 L/h.
3. dm X /dt = y X,S ⋅ qS,f ⋅ cS,f ; m X (t) = m X,0 + y X,S ⋅ qS,f ⋅ cS,f ⋅ t; r X = y X,S ⋅ qS,f ⋅
cS,f /(m X,0 + y X,S ⋅ qS,f ⋅ cS,f ⋅ t); μ drops slowly and hyperbolically to zero.
4. The theoretical final concentration is c X,end,theor = y X,S ⋅ cS,f . Due to lower concen-
trations in the inoculum this can be reached only in a reactor of infinite volume.
The problem will be overcome in a production train, where the content of one re-
actor is used as inoculum for the next one. During cyclic harvesting and filling, the
biomass concentration will come closer and closer to the theoretical maximum.
8 Microalgae – the solar cell factory
The dramatic increase of the world population from seven billion in 2016 to an an-
ticipated eight billion in 2030 is a global challenge. This makes an increase of food
production of 40%, energy of 50%, and raw materials of 100% urgently necessary.
Economics based on fossil resources has reached its limits and even now cannot meet
the demands of climate and environment protection. Global carbon dioxide emissions
are stable at about 40 billion tons per year but are much too high to mitigate global
warming. In the Paris climate agreement of 2015 a consensus was reached to reduce
greenhouse gases and environmental destruction. But how can that be done and what
is the contribution of process engineering?
The sun is the measure of all things. This was already known in ancient cultures
that worshiped the sun as a god or goddess. Today humans gain sun energy by pho-
tovoltaics, while wind and wave energy are also driven by the sun. However, the over-
whelming gift from the sun is food and feed from terrestrial plants. While sun energy as
such is not limited, arable land cannot be further enlarged mainly due to water short-
age. This is the point where microalgae come in. Microalgae have been recognized
as biomass of the ‘third generation’. This view comprises the usage as food, feed, or
fuel. In this chapter we investigate the necessary technology to produce microalgae as
alternative biomass. These photobiotechnological processes are so different from het-
erotrophic bioprocesses and the anticipated production volumes are so huge that mi-
croalgae deserve their own chapter. Here we follow the same read thread as generally
followed in this book, from microorganisms and products through kinetics, media,
bioreactors, and finally integration into a larger context.

8.1 Becoming curious – current research promises

The potential of microalgae has already found public interest. Barack Obama’s 2015
clean power plan talks about “a huge win for algae.” Different media are full of presen-
tations of large cultivation facilities and small start-ups dealing with microalgae. Here
we collect some of the repeated statements often made and append some spontaneous
comments.
– Microalgae are the only way to achieve renewable biomass! Of course, microalgae
offer big opportunities. Some other options are macroalgae or saltwater resistant
vascular plants. With respect to fuels, power-to-fuel technology is a competitor.
– Microalgae grow five times as fast as terrestrial plants! A single algal cell cannot
divide faster than the shoot cells of higher plants. What is meant by this statement
is growth of a culture at given light conditions in an annual average. Plants let
light pass by chance into the soil, suffer from temperature changes, dryness, and
different seasons. This statement of higher productivity will be true only if the

https://doi.org/10.1515/9783110315394-008
 8.2 Choosing from diversity – widely used microalgae for commercial bioprocesses | 183

microalgae are protected and optimally cultivated by taking advantage of their

unicellular habit and growth in suspension.
– Production is possible in arid areas where no other claims exist, as water demand
is minimal! This may be true in closed photobioreactors, but cultivation in open
ponds is accompanied by evaporation.
– Growth takes place in sweet- and saltwater (in contrast to most agricultural crops).
That is obviously true. Saltwater is available in excess. Large production facilities
will be close to coasts, where salt or brackish water is available.
– The whole cell can be used as product without persistent residues! Of course algae
do not have a stem with lignin, which is difficult to recycle. Nevertheless, they have
a cell wall partly consisting of persistent polysaccharides. In other cases we are
interested only in specific compounds.
– Microalgae take up a lot of carbon dioxide! In this point they are not different from
all other plants. What sounds like a big opportunity can turn out to be a technical
challenge.

In the next paragraphs we go into engineering details and try to understand how
the sun, the cells, the reactors, and all other necessary items interact, how effective
microalgal bioprocess engineering is, and how current limitations could possibly be
overcome.

8.2 Choosing from diversity – widely used microalgae

for commercial bioprocesses

Microalgae are photosynthetic organisms without roots, stems or leaves (Latin ‘alga’
= beach, seaweed). These photosynthetic microorganisms are ubiquitous (aquatic
and/or terrestrial), however individual species occupy specific habitats so that some
of them are motile like animals, whereas others are suspended in water or located
on soils (trees or animals), or even in symbiotic relationships with other organisms
(e.g., corals, lichens, flat worms, or other animals). This evolutionary specialization
results in further morphological diversification of microalgae strains, which may be
unicellular, multicellular, round, oval, or filamentous. Microalgae are not a defined
taxonomic group but belong to several taxa. So the term ‘microalgae’ describes more
a habit like bush or tree than a phylogenetically uniform classification. Major groups
like the prokaryotic cyanobacteria and the eukaryotic microalgae (e.g., from the phyla
Chlorophyceae, Trebouxiophyceae, Bacillariophyceae, and Eustigmatophyceae) have
to be distinguished. While cyanobacteria (blue-green algae) have a cell structure very
similar to the prokaryotic bacteria, the individual eukaryotic microalgae cells typi-
cally contain a nucleus, one or more chloroplasts, some mitochondria, Golgi bodies,
endoplasmic reticulum, and other organelles embedded in the cytoplasm. Indepen-
184 | 8 Microalgae – the solar cell factory

dent from cyanobacterial or eukaryotic origin, and multi- or single-cell structure, the
size of individual cells is usually in the range of 3–20 µm.
Considering the large biological diversity and the new developments in the field of
genetic engineering, microalgae are foreseen as one of the most promising sources for
new products and solutions. Between 200,000 and several million species of microal-
gae are expected to exist on Earth, from which approximately between 40,000 and
60,000 species have been identified so far, the chemical composition of only a few
hundred species has been investigated in detail, and nowadays less than 15 strains
are used for cultivation on an industrial scale. Biological fundamentals of the most
relevant microalgae used for commercial purposes will be briefly described in the fol-
lowing paragraphs showing/explaining the reasons why these microalgae can be suc-
cessfully cultivated on a commercial scale.
Cyanobacteria represent the oldest algae (at least 3.8 billion years old), and thus
they are presumably responsible for oxygen evolution on Earth. These prokaryotes
with a gram-negative cell walls are extraordinarily robust and efficient photosynthetic
organisms. Nevertheless, cyanobacteria are very diverse, so for example some species
produce toxins that can taint and poison drinking water (e.g., Microcystis). Others are
harmless or even ecologically indispensable (e.g., oceanic picoplankton Prochlorococ-
cus is the most abundant organism on the planet). Industrially, the most important
and well known species are Arthrospira (Spirulina), Synechocooccus, and Synechocys-
tis.
Spirulina (Figure 8.1) is commonly used as a commercial name for the two indus-
trial relevant strains Arthrospira platensis and Arthrospira maxima, which are both
multicellular, filamentous cyanobacteria (blue-green algae) composed of cylindrical
cells arranged in helicoidal trichome helices. Arthrospira has a very broad physiolog-
ical plasticity when it comes to medium/water salinity conditions, so it can live in
freshwater (salinity < 2.5 g L−1 ) but is the dominant flora in water bodies with high
carbonate and bicarbonate alkalinity and high pH (9–10.5). Generally, temperatures

Fig. 8.1: (a) Macroscopic view of Arthrospira biomass in a lab culture; notable is the filamentous,
felt-like structure (© V. Klassen). (b) Microscopic structure showing a filament and singe cells sepa-
rated by the cell walls (© Science Photo).
 8.2 Choosing from diversity – widely used microalgae for commercial bioprocesses | 185

in the range of 35–38 °C are regarded as optimal for growth and photosynthetic per-
formance of Spirulina. Its ability to grow rapidly under natural conditions make it
a promising source of food and food supplements. The natural blue colorant phyco-
cyanin, along with other biomass components, is an important product of Spirulina.
This and the fact that Spirulina has an easily digestible cell wall have led to the rapid
development of industrial plants for large scale biomass production.
From the wide variety of eukaryotic microalgae, biotechnology currently fo-
cuses on the green algae Chlorophyta (Chlorella, Dunaliella, Haematococcus, Chlamy-
domonas), red and brown algae as well as some diatoms. The green microalgae are
phylogenetically the closest to the higher plants and include a wide range of organ-
isms with very different morphologies ranging from micro- to macroscopic forms. The
robust cell wall of most green algae improves the mechanical stability of the cell, while
reducing the shear sensitivity to mixing or centrifugation, but also complicates the
purification of the inner cell products and digestibility when used as food. Chlorella
vulgaris is a unicellular freshwater microalgae, naturally living in both aquatic and
terrestrial habitats. C. vulgaris cells are round, nonmotile and comparably small with
the size ranging from 4 to 10 µm in diameter. The reproduction (optimal temperature
30–34 °C) is asexual by autospore (2–16) production from one mother cell, which is
divided into three stages: growth (increase in cell size), ripening (mitosis), and di-
vision (release of daughter cells). Chlorella species possess very rigid cell walls, the
structure of which can vary greatly among species. Many Chlorella species are capable
of mixotrophic or even heterotrophic growth on organic substances such as glucose,
acetate, and glycerol.
Haematococcus pluvialis (Chlorophyceae) (Figure 8.2) is a unicellular freshwa-
ter microalgae from the family Haematococcaceae and is industrially used because
of its capability to produce high amounts (up to 4% of dry weight [DW]) of astaxan-
thin (3,3‘-dihydroxy- ß,ß -carotene-4,4‘-dione), which is used as a human nutraceu-
tical and a feed additive in aquaculture. Its cell biology is more differentiated com-

Fig. 8.2: Microscopic picture of Haematococcus; (a) the green mobile stage (macrozooid); (b) produc-
tion stage of astaxanthin (metatocyst). (© Both pictures Science Photo).
186 | 8 Microalgae – the solar cell factory

pared to Chlorella and Spirulina, so for example there are four types of cells iden-
tifiable in the life cycle of H. pluvialis: 1.) macrozooids (vegetative flagellated green
cells), 2.) palmella (nonflagellated green cells), 3.) hematocysts (big red cells with
thick and rigid cell wall) and 4.) microzooids (small flagellated gametes). The op-
timal temperature for vegetative cell (macrozooid) growth/proliferation is between
25 °C and 28 °C. Higher temperatures (e.g., 30 °C) induce cell division inhibition and
lead to an increase in the cell size (palmella) and astaxanthin concentration (hemato-
cysts). Dunaliella salina (Chlorophyceae) is, as the name reveals, a saltwater flagellate,
which grows under very high salt concentrations (growth over a salinity range of 10%
to 35% (w/v), optimum at 22%) and accumulates up to 16% of dry weight (DW) of an
industrially useful pigment ß-carotene (optimal salinity >30% (w/v). The exact func-
tion of these high levels of ß-carotene is not known, but a photoprotective effect is as-
sumed. The optimum growth temperature is dependent on the particular strain, rang-
ing from 21 °C to 40 °C. Thus, the physiological properties of D. salina (high salt and
temperature tolerance, ß-carotene productivity) have led to a variety of different com-
mercial open pond cultivation plants in certain countries (Australia, Israel, India, and
China). Chlamydomonas reinhardtii (Chlorophyceae) is a freshwater microalgae that
can be found in aquatic as well as terrestrial habitats. C. reinhardtii has been studied
intensively since the nineteenth century, so that it is regarded as the most important
and best investigated model microalgae species today. C. reinhardtii genomes (nuclear,
chloroplast, and mitochondrial) are sequenced and it can be genetically modified by
diverse transformation methods. Future improvements of biomass and other valuable
component yields, as well as the reduction of the production costs, will require a thor-
ough understanding of microalgae biology and the transfer of this knowledge to the
industrial production systems.

8.3 Physiological principles – the unique features

of the microalgal cell

To understand the physiological background of photosynthesis, growth and product

formation, and the interaction with the environment, the basic morphological struc-
ture of a eukaryotic algal cell is shown in Figure 8.3 and explained in the next para-
graphs.
Next to the nucleus the cells possess at least one chloroplast and one mitochon-
drion embedded in the cytoplasm. The cell wall and the cell membrane surround the
intracellular components. Further organelles can be found such as the Golgi appa-
ratus (can be more than one), endoplasmic reticulum, and vacuoles. The pyrenoid
(can be also found in higher counts) is located inside the cell’s chloroplast. It con-
tains the enzyme ribulose-1,5-bisphosphat-carboxylase/-oxygenase (RuBisCO), which
is responsible for carbon fixation. The pyrenoid is associated with the presence of car-
 8.3 Physiological principles – the unique features of the microalgal cell | 187

Microtubules

Flagellum

Mitochondrion
Vacuole
Nucleus
Thylakoid stacks
Cytoplasm

Golgi-apparatus

Chloroplast
Starch granule
Cell wall

Cell membrane
Pyrenoid

Fig. 8.3: Morphological structure and principal cell organelles of a eukaryotic algal cell.

bon concentrating mechanisms and starch granules surrounding it. The large chloro-
plast is the most prominent structural feature in a eukaryotic algal cell. It can make up
40% of the cellular volume for phototrophically grown cells. It is defined by a double
membrane (external and internal membrane) and is comprised of several thylakoids.
Chloroplasts, like mitochondria, possess their own DNA. This is attributed to their evo-
lution by endosymbiosis, originating from first photosynthetic cyanobacteria and aer-
obic prokaryotes, respectively. Inside the chloroplast, the circular chloroplast DNA is
contained as well as ribosomes and plastoglobules (lipoproteins as subcompartments
of the chloroplast). Lipid droplets, ribosomes, as well as membrane-bound bodies like
lysosomes, peroxisomes, and glyoxysomes, can also be found in some eukaryotic al-
gae cells. The cellular movement is accomplished, in the case of motile algal species,
with the help of flagella reaching into the cytoplasm, located at the apical side of the
cell. Movement of these flagella is achieved with the attached microtubules. In the
case of the green alga C. reinhardtii, the flagella have a length of 15–20 µm and can be
rebuilt upon loss or damage.
The thylakoid membrane system, surrounded by the stroma, is the carrier of the
photosynthetically active elements. In the case of the green alga C. reinhardtii, the
thylakoid membranes appear either in single form or arranged in stacks of two to ten
discs, but less stacked than the multidisc grana present in higher plants. Moreover,
more complex arrangements like merging and bifurcating discs along the thylakoid
lengths may be found in this alga. A simplified general scheme presenting the inner
structure of the chloroplast is shown in Figure 8.4. According to the theory of sym-
biogenesis the chloroplast represents the former prokaryotic cyanobacterium with a
188 | 8 Microalgae – the solar cell factory

Lamella Thylakoid discs Plastoglobule

Ribosome Intermembrane
Stroma
space
Thylakoid Chloroplast
membrane stroma
Inter-
membrane
space

Thylakoid
Thylakoid
lumen
External membrane Internal
Chloroplast DNA
membrane Thylakoid

Fig. 8.4: Section out of a chloroplast, showing the inner structure of the chloroplast with arrange-
ment of the thylakoid membrane system.

8H+
Chloroplast stroma 2NADP+ 4H+
2NADPH/H+ 3ADP 3ATP

4hν 4hν Fnr

8H+
Fd
membrane
Thylakoid

4e- 4e-
Cyt
LHCII

LHCII

LHCI

LHCI

PSII PSI ATP

b6 f 4e- Synthase
PC
2H2O O2 4H+ 8H+
12H+

Thylakoid lumen

Fig. 8.5: The thylakoid membrane as the site of the light reaction, the primary photosynthetic pro-
cess; the main steps of proton and electron flow are indicated. LHC: light harvesting complex, PS:
photosystem, PQ: plastoquinone, Cytb6f: cytochrome b6f, PC: plastocyanin, Fd: Ferredoxin, and Fnr:
ferredoxin-NADP reductase.

double membrane, while the stroma corresponds to its cytoplasm and the lumen of
the thylakoids to bacterial thylakoids.
The first events during light reactions occur within the thylakoids. These first steps
in the photosynthetic light reaction process are associated with pigment-protein com-
plexes. They can be divided, relating to their function, into the group of light harvest-
ing complexes, photosystems, and cytochromes, playing an important role in electron
transfer. With respect to their function two types of photosystems, namely PSII and
PSI, with their respective harvesting complexes, are distinguished. Their numbers are
related to the time of their discovery. Based on Figure 8.5 we will follow the energy
flow from light capturing over photosynthesis to the first photosynthetic products.
Light harvesting complexes (LHCI, LHCII) are responsible for light energy collec-
tion and the transfer of light excitation energy to the photosystems. Light harvesting
complexes increase the effective absorption cross-section of the photosystems. For the
 8.3 Physiological principles – the unique features of the microalgal cell | 189

purpose of light energy capture LHCs contain about 200 molecules of several different
chlorophylls (Chl a, Chl b, and others) and different carotenoids. Following light ab-
sorption, from a basic view, one electron of a pigment molecule of a light harvesting
complex is shifted to a higher energy level, the excited state. This energy is further
transferred nonradiatively to an adjacent Chl and finally to a reaction center (named
fluorescence resonance energy transfer, FRET, Förster).
Photosystem complexes PSII and PSI contain these reaction centers and consti-
tute the sites of primary chemical energy conversion processes. Photosystems of type
I use FeS clusters (Fe and S containing proteins like ferredoxin, Fd) as final electron ac-
ceptor, whereas type II photosystems make use of quinones (like plastoquinone, PQ)
for that purpose. In higher plants and green algae both types of reactions centers (PSI
and PSII) can be found. Both types of photosystems possess their own core antenna
pigments. However, LHCs contribute the majority (2–4 fold) to the total pigment equip-
ment. Inside the reaction centers is a special central pair of Chl molecules, which are
able to use the collected and transferred excitation energy in a photochemical reac-
tion. In the PSI center is a pair of Chl molecules showing peak absorption at 700 nm.
Therefore, PSI is also termed P700. In contrast, PSII is equipped with a central Chl a
pair with peak absorption at 680 nm and therefore it is referred to as P680.
Both PS and the cytochromes work together to form a linear electron transport
chain, which is outlined in this paragraph. PSII is the first photosystem to be excited
by light energy. The reaction center catalyzes the photolysis of water (water splitting)
including charge separation producing molecular oxygen and four protons from two
water molecules. Oxygen and protons are released into the thylakoid lumen. The
four electrons from this reaction are further transferred via cytochromes to PSI. On
their way, the electrons spend their energy to pump eight additional protons from the
stroma to the lumen against the already existing proton gradient. Finally the electrons
are shifted in PSI to a higher energy level driven again by absorption of an additional
photon. Because of the linear transfer of electrons this pattern is called linear or non-
cyclic electron transfer. The two energy shifts of electrons in PSII and PSI and the
linear electron transfer from PSII to PSI evokes the idea of a lying ‘Z’, which is why
this scheme of energy levels is commonly termed ‘Z scheme’. Now the energy stored
in the proton gradient has to be converted to ATP. This is performed by the enzyme
ATP synthase, which is a pore protein that acts similar to a water turbine, converting
flux energy of protons not into mechanical but into chemical energy. This is actually
phosphorylation of ADP to ATP, where the 12 protons yield three ATP. This is not an
exactly stoichiometric process, so the actual yield can vary and can be smaller. Finally
the gross balance of the light reaction is given as:
4hν
2H2 O + 2NADP+ + 3ADP + 3Pi 󳨀→ O2 + 2NADPH/H+ + 3ATP (8.1)
In cases where the ATP yield in the ATPase is lower or other disturbances occur, the
cells can lead electrons back from the cytochrome to PSII. This is called ‘cyclic electron
transfer’ and increases the ATP/NADPH ratio.
190 | 8 Microalgae – the solar cell factory

Now a second basic process has to follow, which is the fixation of CO2 . This process
is called the dark reaction or more specifically the ‘Calvin cycle’. In plants and eukary-
otic microalgae dark reactions take place inside the chloroplast stroma; for photosyn-
thetic bacteria inside the cytoplasm. For the assimilation of CO2 through reduction of
CO2 , energy in the form of ATP and NADPH as reducing equivalents are necessary, pro-
vided by light reactions of photosynthesis. The cycle can be seen as sequence of three
steps: carbon fixation, reduction, and the regeneration of ribulose-1,5-bisphosphate
(RuBP). The first step is catalyzed by the enzyme ribulose-1,5-bisphosphate carboxy-
lase/oxygenase (RuBisCO). The cycle is initiated by the addition of a CO2 molecule
by covalent bonding to the C5 body ribulose-1,5-bisphosphate (RuBP) resulting in a
C6 molecule. This C6 body is unstable and disintegrates into two molecules of the
C3 body 3-phosphoglycerate (3-PG). During the second step of the cycle, the two 3-
PG molecules are phosphorylated, consuming ATP, to yield 1,3-bisphosphoglycerate
(1,3-BPG). For reduction of 1,3-BPH to glyceraldehyde-3-phosphate (G-3-P), six NADPH
originating from light reactions are needed. For a total of three molecules of CO2 , six
molecules of G3P are produced. One of these C3 bodies leaves the cycle so that equal
amounts of C atoms leave and enter the cycle. The third step of the cycle is character-
ized by the regeneration of RuBP under ATP consumption with the remaining five C3
bodies. Hence, the cycle can be initiated again by addition of CO2 catalyzed by the en-
zyme RuBisCO. Per one molecule glyceraldehyde-3-phosphate (G-3-P) nine ATP and six
NADPH molecules are needed, or per one molecule CO2 two NADPH and three ATP are
consumed. G-3-P is transported into the cytosol and can be used as a precursor for the
synthesis of glucose and starch. Glucose is formed in the cytosol by gluconeogenesis,
the reversal of the first four steps of glycolysis. The remaining G-3-P can be converted
to starch, which is often stored around the pyrenoid in the form of starch granules or
sheets visible in light microscopic images. The gross formula for the dark reaction is
therefore:

3CO2 + 6NADPH/H+ + 9ATP → G3P + 6NADP+ + 9ADP + 8Pi (8.2)

Note that inside the mitochondria the redox as well as the energy balance is completely
closed. The well known gross equation for the whole photosynthesis is finally:
24hν
6CO2 + 12H2 O 󳨀→ C6 H12 O6 + 6O2 + 6H2 O (8.3)

To fuel growth processes inside the cytoplasm, ATP and NADH have to be produced in
analogy to a heterotrophic organism feeding on the produced starch. A final overview
over metabolic pathways and their spatial organization is given in Figure 8.6.
 8.4 Light – the little bit different energy source | 191

Cytoplasm
Mitochondrion
CO2 CO2
O2

Thylakoid lumen
CO2
ATP RuBisCO Respiraon
Chloroplast stroma
ADP
NADPH Dark
NADP O2 PYR ATP
Thylakoid disc reacon
Light
reacon
Chloroplast Starch
H2 O O2 Glucose

G3P
Biomass

O2 Pyrenoid Starch granule

Fig. 8.6: Overview of the main metabolic pathways in phototrophic cells including light and dark
reactions, respiration, and growth in their respective compartments.

8.4 Light – the little bit different energy source

Microalgae as photoautotrophic microorganisms need light as a primary energy

source. Light is a form of electromagnetic radiation in a given spectral range and
propagating in a defined direction. Light is physically measured as an energy flux
called light intensity. More specifically for technical application and in the context of
photobiotechnology, the light energy flowing through a given area per time is called
irradiance I e [W ⋅ m−2 ]. The index ‘e’ recalls the energy flux. To get an idea of the
primary irradiance from the sun and its spectral distribution we look at measure-
ments of sunlight given in Figure 8.7. In this graph the spectral irradiance I e,spec (λ)
[W ⋅ m−2 ⋅ μm−1 ] is plotted over wavelength, meaning the contribution of light to the
energy flux is divided in small slices (1 µm) of wavelength λ. The irradiance in a given
wavelength range of interest is therefore given as:
λ max

I e,Range = ∫ I e,spec (λ) ⋅ dλ (8.4)

λ min

The light energy EA [MJ ⋅ m−2 ] in a spectral range falling on a given ground area in a
given time interval, e.g., a day or a year, is calculated as:
t2

EA,Range = ∫ I e,Range ⋅ dt (8.5)

t1
192 | 8 Microalgae – the solar cell factory

2500
Extraterrestrial radiaon (above the atmosphere)
Terrestrial radiaon
2000
O3
Solar irradiance [W⋅m-2⋅μm-1]

O2
1500
H2O

1000 H2O

H2O
CO2
500

0
Ultraviolet
PAR Infrared (IR)
(UV)
0 500 1000 1500 2000 2500

Wavelength [nm]

Fig. 8.7: Spectral irradiance of sunlight; absorption of light by the atmosphere and especially the
marked singularities are evoked especially by water (clouds), oxygen, ozone, and carbon dioxide
(greenhouse effect!).

The spectrum of light is classified in different wavelength ranges. The infrared range
(IR) makes about 43% of the total solar irradiance (energy flux). Its role in photo-
bioreactor operation is addressed later in this chapter. The ultraviolet range (UV)
contributes 7%. The different wavelength ranges of light have different physiological
impacts. Only the photosynthetically active radiation (PAR) from 400 nm to 700 nm,
meaning 50% of the solar energy, is used by microalgae for photosynthesis. Note that
visible light is nearly of the same range. IR is not used. Biologists think that since
water is a strong IR filter, photosynthesis was ‘invented’ by nature in water environ-
ment. The values are given understanding a vertical position of the sun. However, the
real irradiation depends on the time of day and the latitude of our position. Further-
more, we have to decide in which direction to measure light intensity. As an example
different measurements for Las Vegas are plotted in Figure 8.8.
‘Direct normal irradiance’ is measured directly towards the sun. This is what may
be important for our personal feeling or for photovoltaics in cases when the modules
can be turned towards the sun. ‘Global horizontal irradiance’ is the radiation hitting a
horizontal plane. This is indeed the measure we need to assess light impact on photo-
bioreactors mounted on ground surfaces. Not all of the photons find their way directly
from the sun to a given spot under investigation, but are scattered by clouds or re-
flected from mountains, buildings or trees. This fraction of radiation is summarized
as ‘diffuse horizontal irradiance’.
 8.4 Light – the little bit different energy source | 193

2750
1200 Global horizontal irradiance
2500
Direct normal irradiance
Diffuse horizontal irradiance (calculated) 2250
1000
Solar irradiance [W/m ]
2

2000

PAR range [μE⋅m ⋅s ]

-1
-2
800 1750

1500
600 1250

1000
400
750

500
200
250

0 0
0 2 4 6 8 10 12 14 16 18 20 22 24

Local me [h]

Fig. 8.8: Real data of irradiance in Las Vegas, 21 June 2011, a slightly cloudy day, with different ways
to characterize spatial aspects of irradiance.

From the photosynthesis paragraph we already now that the quantum efficiency of
photosynthesis is wavelength independent. This justifies describing light also over
photon flux density (PFD), which is a useful and widely used variable in this con-
text and denoted here as I hν [μE ⋅ m−2 ⋅ s−1 ]. The index ‘hν’ indicates that the variable
means the number of photons impinging on a given surface per s. The abbreviation
‘μE’ is a reference to Albert Einstein and means 1 micromole of photons. The photo-
synthetically active photon flux density is abbreviated as PPFD and here denoted as
I hν,PAR.
On a technical level energetic considerations are in the foreground, so we need
the relation between irradiance on an energy basis and on a photon flux density basis.
Each photon represents a given photon energy as given by Max Planck as:
c
E hν (f ) = h ⋅ f and E hν (λ) = h ⋅ (8.6)
λ
E hν : energy amount of a single photon [J]; h: Planck constant (6.626 ⋅ 10−34 J ⋅ s); c:
speed of light (2.998 ⋅ 108 m ⋅ s−1 ); λ: wavelength of the photon [m]; and f: frequency
of the light wave [s−1 ] [Hz]. ‘Blue’ photons e.g., have a higher energy content (400 nm,
4.97 ⋅ 10−19 J) than ‘red’ photons (700 nm, 2.84 ⋅ 10−19 J). The mean value of photon
energy in the PAR range (λmean = 550 nm, linear in λ!) is 3.61 ⋅ 10−19 J. The energy
content of 1 E of a 550 nm photon is consequently NA ⋅ 10−6 ⋅ E hν = 6.02 ⋅ 1023 mol−1 ⋅
3.61 ⋅ 10−19 J = 217.3 kJ ⋅ mol−1 .
194 | 8 Microalgae – the solar cell factory

Tab. 8.1: Energetic values of ground radiation.

Location Latitude Annual Annual Power Intensity an- Max Inten-

energy energy average nual average sity noon
measured PAR PAR PAR PAR
[∘ N] [MJ ⋅ m−2 ] [MJ ⋅ m−2 ] [W ⋅ m−2 ] [μE ⋅ m−2 ⋅ s−1 ] [μE ⋅ m−2 ⋅ s−1 ]
Windhoek −22.6 8755 4027 128 591 2370
Las Vegas 36.1 7190 3320 105 488 2180
Karlsruhe 49.0 3231 1486 47 218 1940

Now the needed relation reads:

λ max

I e,Range = ∫ I hν (λ) ⋅ E hν (λ) ⋅ dλ (8.7)

λ min

Now an overview of the situation at real locations should be taken (Table 8.1).This
gives an idea of how much sun energy can be harnessed.
We will need this table further down to calculate the possible microalgae produc-
tion at different locations. It is somehow astonishing that it is only 2.5 times more sun
energy is available at tropical regions than in higher latitudes. The total annual so-
lar energy impinging on our planet is about 1.5 ⋅ 1018 kWh ⋅ a−1 , while the world en-
ergy demand (2016) is 1.8 ⋅ 1014 kWh ⋅ a−1 . Indeed, there is enough sun energy, nearly
10,000 times more than needed; the problem is that it is highly dispersed over large
areas, one of the principle problems of microalgal biotechnology.

8.5 Media for microalgae – a clear matter of fact

Although microalgae may be more variable with respect to macromolecular compo-

sition, medium design follows the same principles as in the heterotrophic case. Nev-
ertheless, some aspects have to be considered coming from the specific habitats of
microalgae. Concerning different habitats of microalgae one distinguishes commonly
between freshwater and seawater algae. Cyanobacteria, as the prokaryotic first pho-
tosynthetic organisms, have nutrient requirements differing from their eukaryotic de-
scendants and can be found in freshwater and seawater or in terrestrial habitats. The
group of freshwater microalgae is also comprised of species not only distributed in wa-
ter bodies like lakes but species also present in soil water like the microalgae C. rein-
hardtii. Additionally, there are also microalgae species distributed in brackish water
or on surfaces of plants, animals, ground, or rock. Hence, according to their different
habitats, microalgae species can show very different nutritional requirements espe-
cially concerning salinity and osmotic pressure. For industrial microalgae processes
the topic of contamination of reactor systems (either open systems like open ponds,
 8.5 Media for microalgae – a clear matter of fact | 195

raceway ponds, or also closed reactors) can be of importance. Here, highly selective
growth conditions, like a high pH (10–11 for the cyanobacterium Arthrospira platen-
sis for phycocyanin production) and/or a high salt concentration of seawater (for the
green alga Dunaliella salina in β-carotene production) enable an axenic culture or very
little contamination also for open systems in the larger, industrial scale.
To start with, a look at the elemental composition (Table 8.2) is required.

Tab. 8.2: Elemental composition of microalgae and a bacterium for comparison; data are given as
percentages.

C N P S K Mg Ca Fe

E. coli 47.0 11.0 2.5 0.9 0.012 0.08 1.27 0.002

C. reinh. 45.3 9.17 3.16 0.42 0.38 0.4 0.92

With respect to the macroelements C, N, P, and S the composition between the different
species is similar. This is expected as proteins and nucleic acids make up the main
compounds. Storage and carbohydrates can change the level of these elements. Ions
not directly built into macromolecules show some differences.
Now we find out whether these differences affect commonly applied defined
media as listed in Table 8.3. Freshwater algae (e.g., Chlorella or Chlamydomonas)
are often propagated in the TRIS-acetate-phosphate buffer medium, abbreviated as
TAP. In technical media as described here acetate as the organic C source is omitted.
Seawater algae (e.g., Emiliana) are best cultivated in an enriched artificial seawater
(EASW) medium. Cyanobacteria (Spirulina, Synechocystis) need a special cocktail (BG
for ‘blue-green’).
Note that the concentrations are calculated for much lower final biomass concen-
trations (e.g., 3 g/L) compared to heterotrophic cultivations (e.g., 50 g/L). So absolute
amounts in the medium have to be set in relation to biomass produced or to other
elements, preferably N. Nitrogen is mostly applied in the form of nitrate or ammonia,
sulfur in form of sulfate, and phosphorus as hydrogen phosphate or dihydrogen phos-
phate. As on other media these also act as buffers in the concentrations given above.
Concerning metal ions, Fe, followed by Mn, Zn, and Cu ions, are the most abundant
ones in cells. These also have to be supplied by the medium in a suitable concentra-
tion. Also noteworthy is that some microalgae require vitamin supplementation in the
medium, whereas others do not (here EASW medium is complemented with thiamine,
biotin, and vitamin B12 ). In some cases this may be attributed to their original habitat
and potential symbioses with other microorganisms where there is a mutual exchange
of essential metabolites. In lab applications chelating agents and buffers are also ap-
plied.
While media compositions are often optimized by trial and error it is worthwhile
to consider trace elements in more detail as they are strain specific and there is “a
196 | 8 Microalgae – the solar cell factory

Tab. 8.3: Examples of commonly defined freshwater and seawater media for eukaryotic microalgae
and cyanobacteria with ion concentrations; vitamins, organic buffers and chelating agents are left
out for simplicity.

Freshwater, TAP medium Artificial seawater Cyanobacteria, BG11 medium

medium, modified EASW
Ion [mmol/L] [g/L] [mmol/L] [g/L] [mmol/L] [g/L]
NO−3 – – 0.55 0.034 17.65 1.09
NH+4 7.0 0.13 – – – –
K+ 1.64 0.06 8.77 0.342 0.351 0.014
Ca2+ 0.34 0.014 9.14 0.37 0.245 0.01
Cl− 7.75 0.28 483 17 0.51 0.018
Mg2+ 0.41 0.01 47.2 1.15 0.304 0.007
SO2−
4 0.51 0.049 25 2.4 0.305 0.029
HPO2−4 0.62 0.06 – – 0.175 0.017
H2 PO−4 0.40 0.04 0.021 0.002 – –
Na+ 0.3 0.007 416 9.6 23.4 0.54
F− – – 0.066 0.001 – –
SiO2−
3 – – 0.105 0.008 – –
Br− – – 0.725 0.058 – –
Trace [µmol/L] [mg/L] [µmol/L] [mg/L] [µmol/L] [mg/L]
elements
MoO2− 4 – – 0.006 9.8 ⋅ 10−4 1.6 0.26
Mo7 O6− 24 0.89 0.94 – – – –
SeO2−3 – – 0.001 1.27 ⋅ 10−4 – –
Zn2+ 77 5 0.254 0.017 0.77 0.05
Mn2+ 26 1.43 2.42 0.133 9.2 0.51
Fe3+ , Fe2+ 17.98 1.0 6.56 0.366 24.5 1.37
Co2+ 6.7 0.40 0.006 3.35 ⋅ 10−4 0.267 0.016
Cu2+ 6.4 0.41 – – 0.32 0.02
Sr2+ – – 82 7.19 – –
Ni2+ – – 0.006 3.7 ⋅ 10−4 – –
BO3−
3 180 11 372 21.88 46.3 2.72

great deal of leeway” as somebody in industry said. The basic difference between pho-
totrophs and heterotrophs is of course photosynthesis. Chlorophylls contain Mg2+ as
a central ion. Moreover, Mg2+ is an essential cofactor of the enzyme RuBisCo. Conse-
quently, phototrophic cells and media contain a high concentration of magnesium.
Current research disclosed some other examples: in contrast to C. reinhardtii, other
freshwater algae like Chlorella and some marine algae are known to require boron.
Soil, the original environment Chlamydomonas reinhardtii was isolated from, is a
boron poor substrate, which is assumed to be the reason for its lack of B requirement
compared to other algae species. Ionomics studies, carefully investigating ion demand
on the molecular level, revealed that selenium is required by ‘Chlamy’, as it is called
in lab slang, for selenoproteins. On the other hand, it has been shown that it inhibits
 8.6 Kinetics of photobioprocesses | 197

the synthesis of chlorophyll. In general, the elimination of elements not required for
algal growth also reduces the risk of contamination with unwanted organisms, which
might require these substances. Strontium in EASW has to be given in even higher
amounts for Emiliania for coccolith production.
The last point to be addressed is supply with carbon sources. It is also possible
to grow several species of microalgae under heterotrophic or mixotrophic conditions
for different reasons. Here, an organic C source is added to the medium where usually
acetate (‘A’ in TAP) or rarely a combination of glucose, acetate, glycerol, and sucrose is
used. This may enhance growth but organic carbon sources added to a medium might
lead to an overgrowing by bacterial or fungal contamination. Even if heterotrophic
production in closed bioreactors is for the time being cheaper than phototrophic pro-
duction, it is in the sense of applying microalgae to use CO2 and sunlight. For small
scale lab applications inorganic carbon is supplied by NaHCO3 /Na2 CO3 serving addi-
tionally as carbonate buffer. While algae can take up CO2 , some can also use HCO−3 .
Optimal partial pressure of CO2 is pCO2 = 0.5%. Note that this is about ten times
higher than in air, a sign that photosynthesis developed in times when atmospheric
CO2 fraction was much higher. To overcome the diffusion barrier between gas bubbles
and medium, CO2 gas fractions for aeration have to be set between 5% and 10%. This
makes application of off-gas from gas combustion plants or heterotrophic fermenta-
tions possible.

8.6 Kinetics of photobioprocesses – how microalgae make use

of the nonmiscible substrate

After we have understood how light supports growth, we are now aiming towards
a quantification of microalgal response to light. In analogy to Monod for the het-
erotrophic case, we start with having a look at growth during different light intensities.
The specific growth rate shall be measured and plotted against light intensity I hν,PAR,
the photon flux density (PFD) in the PAR range. This yields the so called photosynthe-
sis irradiance curve (PI curve), where a typical example is shown in Figure 8.9. Such
experiments have to be made understanding that all other nutrients including CO2 are
present in excess, and all cells ‘see’ the same light intensity. Unlike an ideally mixed
tank reactor a homogeneous light distribution cannot be reached by ideal mixing but
only by using low biomass concentrations or specific geometries. Timescale is also
an issue. From a biology viewpoint, photosynthetic activity in terms of oxygen pro-
duction is measured on a short term basis in cuvettes. Measured PI curves from batch
experiments is standard, but gives only a snapshot valid during a given time interval.
Microalgae react during long term light exposition by changing pigment furnishings,
which is known as photoacclimation. Continuous cultivation covers adaptation and
acclimatization and gives the most reliable values for anticipated process design.
198 | 8 Microalgae – the solar cell factory

1.2
I II III

1.0
opmum
operang point
0.8
μ [d-1]

0.6 slope measure

for efficiency
0.4

0.2

0.0
0 300 600 900 1200 1500 1800
Ihν >w(yPyV]

0 4 8 12 16 20 24 28 32
Iabs,hν >w(yJyV]

Fig. 8.9: Typical absorption spectrum of Chlorella vulgaris and the contribution of different pig-
ments.

The PI curve looks remarkably different from Monod kinetics. The first linear phase re-
flects the physical mechanism of light absorption. Here the cells are light limited, de-
pending on the passively impinging light energy without involvement of an enzymatic
step. The slope represents the constant efficiency of growth in this intensity range.
Microalgae also need maintenance energy, which is covered by a part of the energy
gained by respiration. This becomes visible by the intercept of the kinetics with the μ-
and I hν,PAR axis. At the compensation point I hν,comp the energy gained by photosyn-
thesis exactly equals the maintenance energy leading to zero growth. The saturation
point I hν,sat is usually in the range of 200 to 400 [μE ⋅ m−2 ⋅ h−1 ], so it can be found far
below a typical value of sunlight even in middle latitudes.
For radiation values above the saturation point a constant photosynthetic activity,
here measured as growth, is observed. Additional light can obviously not be used by
the cells. So a bottleneck further down the metabolic energy flow is assumed. This can
be directly after light absorption in the light reaction e.g., during water splitting, in the
dark reaction e.g., CO2 fixation, or even further down in the metabolism. So we have
the typical situation of Blackman kinetics (see Chapter 4) including two consecutive
metabolic processing steps. Light energy being absorbed but not further processed is
dissipated as fluorescent light and heat. This mechanism is known as nonphotochem-
 8.6 Kinetics of photobioprocesses | 199

ical quenching (NPQ). The maximum specific growth rate μmax depends strongly on
the strain and can reach values of 1 g ⋅ g−1 ⋅ d−1 or even more than 2 g ⋅ g−1 ⋅ d−1 .
Above saturation light intensity Isat μ decreases. This situation is usually avoided
during outdoor cultivation. However, some microalgae produce light protecting pig-
ments like carotenoids only in this intensity range, so it is applied to obtain high
product titers. The inhibition range is kinetically often not well defined and can be
of different mechanisms and respective kinetics.
The quantitative description (8.8) of light kinetics can now derived from what was
said above and can be written down as:

{
{ k μ,hν ⋅ I hν,PAR − μ e for I hν,PAR < I hν,sat
{
μ (I hν,PAR) = { μ max [= k μ,hν ⋅ I hν,PAR − μ e ] for I hν,sat < I hν,PAR < I hν,inhi
{
{ 2⋅k hν,inhi
{ μ max ⋅ k hν,inhi +I hν,PAR [k hν,inhi = I hν,inhi] for I hν,PAR > I hν,inhi
(8.8)
This kinetics is a black box approach similar to formal kinetics for heterotrophs and
should be elaborated in more detail to get a transparent interface to physiological
mechanisms. This includes firstly a transport step. This step is enzymatically enabled
or facilitated in the case of heterotrophic microorganisms. In Chapter 4 (kinetics) we
replaced at this point Monod kinetics μ(cS ) with Michaelis–Menten kinetics rS (cS ).
For phototrophs light absorption is the underlying mechanism for energy intake. Sec-
ondly, substrate is converted to biomass, which is a reaction step. Carbon is allotted
to different metabolic pathways finally building up the cell mass. Due to the stoichio-
metric constrains this process is basically formulated by yield coefficients y X,S for
heterotrophs. Speaking in system theory language this is a transport-reaction system.
Now this structure will be applied to phototrophs.
In phototrophic organisms light as the energy source impinges on the cells and
is passively absorbed. Absorbance depends on biomass concentration, on pigment
furnishing of the cells, and will be linearly dependent from light intensity. While the
absorbed light flux I hν,abs represents the available energy flux for the cells, the value of
interest is then the light flux absorbed by a given amount of cells. This value is given
here as r hν,abs [μE ⋅ g−1 ⋅ s−1 ] with r hν,abs = I hν,abs /c X meaning the biomass specific
number of absorbed photons per time in analogy to the specific substrate uptake rate
for the heterotrophic case. This is named ‘light availability’. Referring to the dual char-
acter of the substrate as energy and carbon source, r hν,abs represents of course only the
energy aspect of the substrate. The easiest formulation for this first step is to assume
that r hν,abs is linear to light intensity I hν,PAR given as photon flux density:

I hν,abs
r hν,abs = = σ X ⋅ I hν,PAR (8.9)
cX
The proportional factor denoted as σ X is the absorption cross-section similarly as in
the Beer–Lambert law here with the dimension [m2 ⋅ g−1 ]. It represents the virtual area
on which 1 g of spread algae would absorb the impinging light completely. Of course
200 | 8 Microalgae – the solar cell factory

0.6 0.7

Chlorella

Absorpon cross secon [m ⋅g ⋅nm ]

-1
Car 0.6
0.5
Chl a

-1
Chl b 0.5

2
0.4
Absorpon [AU]

0.4
0.3
0.3

0.2
0.2

0.1
0.1

0.0 0.0
400 450 500 550 600 650 700

Wavelength [nm]

Fig. 8.10: Idealized photosynthesis irradiation response curve; the different intensity ranges are
marked.

this kinetic parameter depends on cell size or pigment content of the cells as well as
on wavelength. A typical absorption cross-section spectrum is shown in Figure 8.10.
The parameter σ X has to be measured for every algae strain and process condition; a
typical value for Chlorella is 0.3 m2 /g for sunlight spectrum.
The typical shape of the curve reflects mainly the two absorbance peaks of chloro-
phyll (around 380 nm blue and around 680 nm red). The absorbance minimum be-
tween 500 nm and 600 nm is called the green gap. This makes microalgae look green
but prevents them using green light efficiently.
After considering the transport step of light absorption, the reaction step is to be
formulated as rate of photosynthesis (here μ) as a function of photon absorption rate,
sometimes called the ‘photosynthesis efficiency curve’. This curve gives a good insight
into the efficiency with which photons are utilized for biomass formation. Photosyn-
thetic efficiency η X,hν can be read from the curve as r X /r hν,abs for each given light in-
tensity or in integrated form from data as ∆m X / ∫ I hν,abs ⋅ dν.

The kinetic equations for phototrophic growth now look like:

r hν,abs = σ X ⋅ I hν,PAR (8.10a)

{k μ,hν,abs ⋅ r hν,abs − μ e [μ ≤ μmax ]

μ(r hν,abs ) = { (8.10b)
μ other environmental factors
{ max
 8.6 Kinetics of photobioprocesses | 201

Unlike heterotrophs microalgae cannot reduce the amount of absorbed light in cases
were growth is limited e.g., by nitrogen or CO2 availability, maybe only slowly in the
acclimation process.
Besides the physiological model the reactor model equations for a batch process
have to be set up. The basic difference here is the transport term for light, which is
as we already know not miscible. In all particulate points in the suspension another
light intensity has to be applied leading to different specific growth rates along the
light path. That forces us to formulate the reactor equations as spatially distributed.
Light transfer in molecular disperse systems obeys formally the Beer–Lambert law
Equation (8.11a):
I hν (lpath ) = I hν,0 ⋅ e−σ X ⋅c X ⋅l path (8.11a)
Here I hν,0 is the incident light intensity and lpath [m] the light path length. However,
this law has been developed for molecular dispersed systems. In the case of a suspen-
sion scattering also occurs, what is considered in σ X . As forward scattering dominates
and side scattering is symmetrical for flat geometries a similar attenuation curve can
be assumed. However, we cannot be sure that light attenuation is really linear with re-
spect to biomass concentration. Basically, μ also depends on the location on the light
path (Figure 8.11).

800

4
I [μE⋅m-2⋅s-1]

600
μ [h-1]

400
2

200
1

0 0
0 5 10 15 20

Fig. 8.11: Exponential decrease of light intensity along the light path; growth follows the light ac-
cording to kinetics and is in the saturation range in the bright part and in the linear range in the
darker part of the reactor.
202 | 8 Microalgae – the solar cell factory

6 0.6 600
I II III
5 0.5 500

4 0.4 400

ITrans. [μE⋅m-2⋅s-1]
cNO3 [g/L]
cx [g/L]

3 0.3 300

2 0.2 200

1 0.1 100

0 0.0 0

0 1 2 3 4 5 6 7 8
tcult [d]

Fig. 8.12: Three different growth phases – exponential, linear, and limited – can be identified,
marked by three sections.

For now we assume that μ can be formulated as an average value μav valid for the
whole reactor to set up the biomass balance (Equation 8.11b):
dc X
= μ av (l hν,0) ⋅ c X (8.11b)
dt
This is second reactor equation is further elaborated for microalgae in Chapter 12.
Now we will test this approach for the case of the flat cuboid reactor and compare
it with measurement data. The result is shown in Figure 8.12. For simplicity only light
in the not light saturated range is applied.
Initially, for low biomass concentration, all cells experience more or less the same
light intensity, which does not change much for a few hours. This leads to exponen-
tial growth. In the longest middle phase of the cultivation all light is absorbed in the
reactor. Light intensity at the dark side of the reactor with thickness DR is nearly zero.
With increasing biomass the amount of photons absorbed per cell slowly decreases,
so we expect a slowly decreasing μav . The biomass balance can be further elaborated
now as (8.12):
dc X 1 I0 ⋅ η X,hν
= μav ⋅ c X ≈ ⋅ I hν,0 ⋅ η X,hν ⋅ (1 − eDR ⋅σ X ⋅c X ) ≈ (8.12)
dt DR ⋅ σ X ⋅ c X DR ⋅ σ X
This is clearly not exponential growth as biomass increase does not depend on
biomass itself but only on the constant flux of absorbed light. This phase is con-
sequently called the linear growth phase. Note the analogy with constant feeding in
 8.7 Photobioreactors – the interface between algae and sunlight | 203

the fed-batch process for yeast production. It is also interesting to observe that dc X /dt
depends on reactor geometry. The thinner the reactor is the faster the increase in
biomass concentration. This can be understood from the fact that the total biomass
produced depends directly on I0 , but is suspended in a volume dependent on DR .
In the late third phase either nutrients are limited or the increasing dark volume
increases the relative amount of maintenance losses.
Photobioreactors can be operated also in continuous mode. Turbidostat is com-
monly applied as well as ‘luminostat’ in lab reactors, where the transmitted light is
measured and kept constant. Daily harvesting in production scale leads to quasicon-
tinuous process policy. Typically, biomass is harvested in the afternoon to combine
the time of strongest sunlight with the highest biomass concentration on the one hand
and to minimize biomass loss by respiration in the night.
With this knowledge in our baggage we can now start with reactor and process
design.

8.7 Photobioreactors – the interface between algae and sunlight

Photobioreactors are the centerpiece of each photobiotechnological process. They can

be artificially illuminated or, as outdoor reactors, illuminated by the sun. In any case
the physiological requirements of microalgae define the construction concepts of pho-
tobioreactors. So the work of an engineer designing a photoreactor resembles the work
of an architect planning a house for a family. The first step is to ask the family what
they need, and a second step is to consider the environmental conditions and make
compromises. We collected already the necessary information, which is kinetics for
the microalgal needs and light conditions at different sides. The reactor has to trans-
form these conditions into suitable conditions inside the reactor.
In terms of hydrodynamics, a photobioreactor represents a four phase system with
the liquid phase providing water and nutrients. The gaseous phase supplies carbon
dioxide and removes excess oxygen from the system. Eventually, the solid phase is
given by the cells. The fourth interacting phase is the superimposed light radiation
field.

8.7.1 Starting from natural conditions – lakes and open ponds

Microalgae live naturally in lakes or in the sea but also in other wet environments and
are adapted to these conditions. So a start is to mimic these natural environments of
course improving some factors for the algae in view of better productivity. Arthrospira
(brand name Spirulina) is produced e.g., in volcanic lakes in Myanmar or soda lakes
in China. The high pH value protects the culture from predators or competing microal-
gae. Harvesting is performed simply by filter nets, which is possible thanks to the fila-
204 | 8 Microalgae – the solar cell factory

Harvest Feed Paddlewheel

Baffle Flow Baffle

Fig. 8.13: Ground plan of a typical raceway pond; the most important gauges and peripheral devices
are delineated.

mentous growth of these cyanobacteria. Cost are low for this extensive approach but
so is productivity. The reasons are poor mass transfer for CO2 and suitable nutrient
supply. Other problems are seasonal fluctuations and last but not least lack of global
availability.
To tackle these problems open ponds are constructed, especially for better mixing
and CO2 supply, and in general to maintain better and constant environmental con-
ditions. These are usually constructed as recirculation ponds and are recirculated by
paddle wheels. Such ‘raceway ponds’ (Figure 8.13) are available in different configu-
rations. While a single pond can be 100 m long, production plants with many ponds
operated in parallel can cover several hectares. In fact, by far the greatest share of
microalgae production comes from raceway ponds.
The mean flow velocity vf [m/s] follows from the need to prevent sedimentation
and to mix the lower with the higher water layers. In practice, often 0.3 m/s are ad-
justed. The necessary power input as a contribution to costs is calculated as:
−1
PW,pond = 1.59 ⋅ AG ⋅ ρ L ⋅ g ⋅ u 3f ⋅ fM2 ⋅ dh 3 (8.13)

This includes the ground area AG [m2 ], the density of the medium ρ L practically as 1
[kg/L], the gravitational acceleration g [m/s2 ], and the manning factor fM [s ⋅ m−1/3 ]
describing the friction at the walls dependent on roughness between 0.01 and 0.016.
The hydraulic diameter dh for the given geometry is 4 ⋅ w ⋅ h/(w + 2 ⋅ h). The mechanical
energy input calculated in this way is with 0.5 W/m2 or 1 W/m3 ; relatively low. The
energetic efficiency of the paddle wheel (Figure 8.14) itself can make up another 50%,
where the solution is blades dipping vertically into the water.
Now we compile some pros and cons of these ‘open reactors’. Areal productivity
PA is with 10 g ⋅ m−2 ⋅ d−1 ≈ 35 t ⋅ ha−1 ⋅ y−1 = relatively low but at least remarkably
higher than one can get from terrestrial plants. CO2 is applied via bubbles behind
 8.7 Photobioreactors – the interface between algae and sunlight | 205

Fig. 8.14: Picture of a raceway pond; the paddle wheel is a CAD drawing showing that the paddles
nearly reach the ground to create a uniform flow over height and to prevent sedimentation (© a4f).

the paddle wheel but a large portion will evaporate from the free surface to the at-
mosphere. Water also evaporates (up to 10 mm/day of water column) and has to be
replaced, preferably by seawater, not to mention the necessary pumping energy. The
salt content will rise but to an acceptable level e.g., when the pond is operated in
the Mediterranean Sea. In some desert areas groundwater is used, which is of course
against the sustainability idea. At last water evaporation helps keep the temperature
quite constant, rising not more than typically 10 °C above ambient air temperature.
Evaporation can be diminished by spanning plastic foils over the then so called ‘cov-
ered pond’. Contamination with predators (daphnia, paramecium), fungi, or compet-
ing microalgae is an important issue.
While large ponds are around 30 cm deep (more than optimal) and are made sim-
ply by gasketed lacunas, modern concepts employ stainless steel trays only 10 cm deep

Fig. 8.15: Inclined surface reactor; the algae suspension flows down into a collection channel for gas
exchange (© TU Munich).
206 | 8 Microalgae – the solar cell factory

mounted in greenhouses. A special reactor type, known already for some time but at-
tracting new attention, is the ‘inclined surface reactor’ (Figure 8.15). It is based on
raceway ponds but with an inclined tray where recirculation is managed not by pad-
dle wheels but by a volumetric pump. The gravity driven suspension film is in this way
only a few mm thick. What this is good for we will learn in the next paragraph.

8.7.2 Flat plate reactors – the bubble column in an ‘enlightened’ guise

The next step forward is to replace all open surfaces by transparent enclosures to gain
a ‘closed photobioreactor’ and analyze which additional design features are possible
compared to the open reactors. The transparent walls could be made of glass, PVC, PE
or PMMA. The first design aspect is to let as much light as possible into a given reactor
volume VR , meaning that the reactor surface to reactor volume ratio SVR = AR /VR
should be as high as possible. For a rectangular geometry both sides count yielding:
2 ⋅ LR ⋅ HR 2
SVR = = (8.14)
LR ⋅ HR ⋅ DR DR
For e.g., DR = 2 cm SVR = 100 m−1 . For a pond of 25 cm depth the value is only
−1
4 m . This is equivalent to making one dimension, here referred to as thickness DR , as
small as possible. The idea has its equivalence also in Equation (8.15). The result is the
‘flat plate’ reactor (Figure 8.16). During the process biomass concentration can be kept
so high that the ambient light impinging on the reactor and being distributed inside
the reactor serves all algae cells with an optimal amount of photons e.g., 50 µE/g/s.

reactor thickness degassing

0.05 m

reactor height
1.5 m

perforated tubes
gassing

gassing
reactor width
3 m, scalable

Fig. 8.16: The principle of a flat plate photobioreactor with peripheral devices. Mutual shading con-
tributes to light dilution (© KIT).
 8.7 Photobioreactors – the interface between algae and sunlight | 207

Direct normal
I0 Horizontal irradiance irradiance
= Incident light
intensity
VR Reactor volume
cX Cell dry mass concentraon
AR Reactor surface Area

AG Aperture (ground, footprint) area

PG Areal
producvity PR Volumetric producvity

Fig. 8.17: The principle of light dilution of ambient light on a large reactor surface and different char-
acteristic values for design and evaluation of photobioreactors.

To mount such a flat plate horizontally on the ground to create a water body similar
to an open pond is not a good idea. The direct sunlight is much stronger than the al-
gae need, at least in the upper layer. Energy loss by NPQ or even light inhibition will
be the result, which is an additional problem in open ponds. The second aspect is to
dilute the sunlight to values near the linear range of the PI curve. This ‘light dilution’
can be obtained by mounting the plates vertically as shown in Figure 8.17. Decisive for
overall productivity is the vertical light component I hν,horiz (global horizontal irradi-
ance), but physiologically active and penetrating the reactor is only the component
I hν,norm perpendicular (normal) to the reactor surface. The light balance is therefore
I hν,horiz ⋅ AG = I hν,norm ⋅ AR .
To summarize this second design aspect in a formula it can be stated that the
reactor surface to reactor footprint (ground) area SFR = AR /AG should be as high as
possible. For a flat plate reactor installation (Figure 8.18) both transparent sides count,

Fig. 8.18: Arrangement of flat plates, the so called ‘green wall panel’, where plastic bags are
mounted between wire fences for stabilization.
208 | 8 Microalgae – the solar cell factory

making:
2 ⋅ HR ⋅ LR 2 ⋅ HR
SFR = = (8.15)
LR ⋅ dR dR
Here dR is the distance between two reactor panels. Mutual shading of reactor panels
and reflection on the surface contribute to light dilution. In technical installations a
factor of up to 10 is realistic. A pond has a value of only 1. Vascular plants like trees
also use this principle. Leaves are very thin but span a high total surface of leaves
(e.g., 2.500 m2 ) over the ground area shaded by the tree (e.g., 100 m2 ). A lime tree
can outnumber a value of SFR = 20 (in botany only one side of a leave is counted). The
two aspects shown above mean also that the fluid volume per ground area has to be
low (10 L/m2 –50 L/m2 ). This is sometimes not understood intuitively, as ‘the medium
is the reaction volume’ is a rule of thumb in reaction engineering. Here the algae cells
make up the real reaction volume.
Mass and mechanical energy transfer in flat plate reactors is ensured by bubble
aeration from a sparger similarly as in bubble columns. As gas transfer is essentially
lower e.g., by a factor 100 than in heterotrophic processes (lower μ, lower c X ), flat
plates are operated in the laminar regime without bubble dispersion and coalescence.
Typical aeration rates are 0.05 to 0.15 vvm. This is with e.g., 50 W/m3 or 1 W/m2 , a con-
siderable part of the chemical energy produced as algae biomass. A careful control of
the aeration rate according to the needs, e.g., lowering at low light conditions or dur-
ing the night, can reduce the costs for energy transfer. Fluctuating CO2 consumption
due to light fluctuations is usually compensated by controlling the CO2 fraction in the
gas.
To reduce pressure loss and therefore energy expenditure the height of the reactor
can be lowered and the number of plates per ground area accordingly increased. This
also reduces material expenditure as the lower hydrostatic pressure at the bottom of
the reactors allows for application of thinner reactor materials. This reflects a current
trend in photobioreactor design.

8.7.3 Tubular reactors

To further increase SVR and SFR employment of transparent tubes is a geometrical

means (Figure 8.19). They are installed horizontally with the length axis being the
length axis of the reactor. The length dimension LR can be several hundred meters.
Several tubes ntube are mounted above each other and are connected by bends in this
way forming so called ‘fences’. Tubular reactors belong to the largest closed reactors.
Light capturing given as SVR results in (8.16):
ntube ⋅ LR ⋅ 2 ⋅ π ⋅ RR 2
SVR = = (8.16)
ntube ⋅ LR ⋅ π ⋅ RR 2 R R

This is quite similar to the flat plate reactor. Especially for light distribution there is an
advantage. Light can penetrate the tube in two perpendicular (horizontal and vertical)
 8.7 Photobioreactors – the interface between algae and sunlight | 209

installaon fence height

length 100m 3m

degassing

gassing tube diameter

5 cm

Fig. 8.19: Principle of tubular reactors with peripheral devices.

directions. This leads to a homogeneous light distribution as exponential light atten-

uation (I(r) ∼ eR−r ) is partially compensated by the ‘focus effect’. This term describes
light beams approaching the suspension in a radial direction coming closer and closer
to each other. This leads to a virtually higher local irradiance, as the photons penetrate
a decreasing circumferential surface area closer to the axis (A(r) ∼ r−1 ).
Light dilution over the surface is another big advantage as interpretation of Equa-
tion (8.17) shows:
ntube ⋅ LR ⋅ 2 ⋅ π ⋅ RR ntube ⋅ 2 ⋅ π ⋅ RR
SFR = = (8.17)
LR ⋅ dR dR
The number of tubes corresponds to the height of a flat panel, but the circumference
term (π ⋅ RR ) increases the value remarkably.
Tubular reactors are operated as plug flow reactors with a flow velocity between
0.2 and 0.5 m/s. Cycle times are of course much smaller than specific growth rates, so
with respect to biomass they behave like ideally mixed tank reactors.
To operate tubular reactors an external pressure difference has to be applied. This
is done either by centrifugal pumps or via the airlift principle, where the reactor it-
self is the downcomer. The pressure difference can be calculated according Hagen–
Poiseuille equation (8.18):
qL ⋅ LR ⋅ 8 ⋅ ηmed vL ⋅ LR ⋅ 8 ⋅ ηmed
∆p = = with qL = π ⋅ vL ⋅ R2R (8.18)
4 ⋅ R4R R2R
Here ηmed is the dynamic viscosity of the medium (ηH20 = 0.9 Pa ⋅ s at 25 °C) and vL
the average medium velocity. What looks a bit frightening is that the radius in the
denominator scales with a power of 4 is directly opposite to the intention of making
it small. But to drive the volume flow through a smaller tube is not necessary, but to
keep the velocity constant. This reduces the problem to a power of 2. Nevertheless, the
high pressure difference has an impact on the mechanical stability of the tubes and
also on the microalgae experiencing a pressure jump during each cycle.
210 | 8 Microalgae – the solar cell factory

Fig. 8.20: Large scale installation of tubular reactors in parallel double fences, making bending
easier (© a4f).

The same holds for the power input (8.19).:

8 ⋅ ηmed ⋅ LR ⋅ q2L
Ptube = = 8 ⋅ ηmed ⋅ LR ⋅ π ⋅ v2L with qL = π ⋅ vL ⋅ R2R (8.19)
π ⋅ R4R
At constant liquid velocity power input is constant for different tube radii, but volu-
metric power input increases with decreasing radii. In practice, tube diameters of 4 cm
are commonly applied leading to a power input of more than 100 W/m3 (Figure 8.20).
Gas transfer is foreseen by agitated vessels coupled into the liquid cycle. During
their way through the reactor the microalgae take up CO2 and produce O2 . This leads
to pH, pCO2 and pO2 gradients. When all dissolved CO2 is used up the medium must
again reach the gas exchange vessel at the latest stage. This is usually after about two
minutes. Besides the pressure drop along the longitudinal axis this limits the lengths
of tubular reactors. A bit of help comes from making gas bubbles drift along the tube.
This is the price we have to pay when taking two spatial coordinates for light transfer.

8.7.4 New developments and ideas

Flat plates can be pronounced as tall bags or long bags submerged in a ‘water bed’
(may also be open seas) to increase mechanical stability and reduce the impact of
heat accumulation. Better flow patterns or light conducting structures (Subitec) for
better light distribution are other trends. Even the horizontal reactor (KIT) comes back
but with corrugated surfaces to cope with the basic parameters mentioned above. The
sun energy ends up nearly completely as heat inside the suspension. This last item
is besides contamination one of the main problems in outdoor cultivation. Modern
 8.8 Products and processes | 211

concepts go away from these basic types. Horizontally arranged reactors would offer
advantages like low pressure and no need for racks. However, light has to be guided
into the suspension e.g., by corrugated surfaces or light conductors. Further, such con-
cepts allow the collection of IR outside the photoactive volume thus decreasing heat
uptake and further the application of ‘transparent’ photovoltaics.

8.8 Products and processes

Microalga biomass is generally composed – like most microbial biomass – from the
main components protein, carbohydrates, lipids, and as a specialty, pigments. So their
nutritional value can be equated with fruits or seeds of higher plants. The difference
with respect to production is clearly that the cells in the suspension represents the
whole biomass and not only a part of the harvest. Furthermore, the fraction of the dif-
ferent compounds can be shifted in wide ranges according to process strategy. Fur-
thermore, microalgae biomass represents a valuable source of a balanced mineral
content (e.g., Na, K, Ca, Mg, Fe, Zn, and trace minerals) and all essential vitamins
(e.g., A, B1, B2, B6, B12, C, E, nicotinate, biotin, folic acid, and pantothenic acid). This
makes microalgae as whole cell preparations, keeping the high value of the balanced
constituents, interesting for commercialization as nutraceuticals. Specific extracts are
commercialized for their different bioactivity e.g., as antioxidants.

8.8.1 Products for food supplements, feed and aquaculture

Microalgae deliver in general similar products as terrestrial plants but offer additional
products as well (Figure 8.21). Firstly, we look at total biomass without any product
extraction.
Chlorella and Spirulina (Arthrospira platensis), both possessing the GRAS (gen-
erally recognized as safe) status, are traditionally used as food supplements and cur-
rently dominate the fast developing market of microalgae based health food produc-
tion. The production volume increased for Spirulina/Chlorella in 2007 from 5,500 t
(US$1.25 billion) to about 15,000 t (US$6 billion) in 2014, 70% being Spirulina. How-
ever, other microalgae species appear also to arise slowly within the food and feed
market. One limitation in bringing interesting new species into the market is lack of
scientific insights relating to nutritional value or unknown side effects, making a so-
phisticated approval process necessary, while Spirulina and Chlorella already possess
the GRAS status, because they are traditionally used. Another cyanobacterium, Aph-
anizomenon flos-aquae with brand names like AFA, has currently made a sensation
in the USA. A general concern in blue-green algae is their ability to produce toxins
(Figure 8.22)
212 | 8 Microalgae – the solar cell factory

Hydrogen Rec. Proteins

Oxygen Unsaturated fay acids
Food Biodiesel
Pigments
Bulk chemicals
Anoxidants
Vitamins
Pharmaceucals Starch
Oil Food addives
Proteins Bioethanol
Light Pigments

Poly-
saccharides
Food supplement
Aquaculture
CO2 New
Feed
biomass
H2O Methane
Minerals

Fig. 8.21: The product range of microalgae.

Fig. 8.22: Microalgae biomass as food supplement are sometimes categorized as ‘superfoods’ and
require a corresponding product presentation and marketing (© bio-compete).

Microalgae for food use are usually produced in raceway ponds. Chlorella has a strong
cell wall, which has to be disrupted before further processing to increase bioavailabil-
ity. Whole cells as dietary supplements are commonly marketed as dry powder, com-
pressed pastilles (green or orange color) or as capsules. It should not be concealed that
current market prices of 10–30 US$/kg are too high to allow for a high enough daily
diet.
In general, proteins represent the most abundant compounds in microalgae
biomass, containing all essential amino acids and also a variety of other bioactive
compounds (e.g., antioxidants, PUFAs, bioaccessible trace elements) and can be used
as a whole for nutraceuticals and feeds. Furthermore, microalgae are important feed
additives for aquaculture animals (fish, shrimp) and are essential feed either for direct
 8.8 Products and processes | 213

larval nutrition (e.g., mollusks or shrimp) or indirect as food for the live prey fed to
small larvae fish (about 75% of produced amounts are used as food/feed or dietary
supplement). Microalgae are the basis of the marine food chain. They are the feed for
the above mentioned animals and especially for krill in the oceans. Small fish feed on
small animals, big fish chase small fish, and finally humans are the predator of big fish.
Edible fish are usually carnivores and get PUFA and carotenoids from this food chain.
So in aquaculture they have to be fed by fishmeal, which is no longer available due to
overfishing the oceans. Direct cultivation of microalgae to supplement the fish feed is
the emerging alternative. Otherwise fish will not be the healthy food it used to be.
Many areas on our planet like parts of Africa or Europe are protein deficit areas. As
long as people can afford it, they buy soya beans e.g., in South America, at the cost of
rainforests and the local population. This is reason enough to think about microalgae
as a protein and vitamin source not only as a food supplement but in large scale. Pro-
tein content in microalgae is up to 50–70% of BDM depending on species and culture
conditions, containing all essential amino acids. The nutritional profile with respect
to amino acids is comparable to traditional food like meat, egg, milk, or soya. The
most abundant proteins in living nature are actually proteins from light harvesting
complexes and RuBisCO; both fractions can be up to 30% of the whole protein. Soy
protein is traded with about 2,000 US$/ton, while microalgae protein is not available
under 10,000 US$/ton. This is a strong incentive for further technology development
also with a view to applicability in countries with limited infrastructure.

8.8.2 High value products

Lipids are essential for all living organisms primary because all cell membranes are
built up from polar lipids. Polar lipids consist mainly of phospholipids and glycol-
ipids. Algae are here no exception but are of unique advantage insofar as they pos-
sess large amounts of membranes in the thylakoids and form especially high amounts
of polyunsaturated fatty acids (PUFAs). Humans lack the ability to introduce double
bonds in fatty acids beyond carbon ‘9’ (Figure 8.23). Very valuable as functional food
are linoleic acid, α-linolenic acid, eicosapentaenoic acid (EPA, 20:5ω-3), and docosa-
hexaenoic acid (DHA, 22:6ω-3).

6 ω3
5 7
3 8 ω1
HO 1 9 ω6

Fig. 8.23: Eicosapentaenoic acid as an example of a merchantable product; numbers (left side of
molecule) represent the regular carbon counting while counting with respect to double bounds
starts from the end (right side of molecule). Single bonds can rotate enabling different spatial struc-
tures of the molecule.
214 | 8 Microalgae – the solar cell factory

EPA and DHA can be found predominately in the polar lipid fraction of marine mi-
croalgae species in particular, and are important for human health. Enhanced DHA
intake e.g., has a positive cardioprotective effect on adults and may improve infant
cognitive performance and enhance visual acuity. The reason why especially marine
species contain higher PUFA amounts of is not clear but low temperature can stimu-
late EPA concentration increase. While for industrial EPA production strains of Nan-
nochloropsis or Phaeodactylum (Rhodophyta) are employed, DHA is produced e.g., by
Schizochytrium. Current screening programs aim at finding strains that are good pro-
ducers of both PUFAs (e.g., Pavlova lutheri). Essential fatty acids are sold predomi-
nately as oil capsules or are worked into functional food. The natural supply is via
some plant oils or via sea fish (or extracts from fish liver), again a hint to the marine
food chain from algae to fish.
Another group of lipids found in microalgae are neutral lipids (e.g., triacylglyc-
erol, TAG), which accumulate in the cells up to 60% DW as storage lipids. TAGs are in-
teresting as future solar fuels, however comparably high cultivation, harvesting, and
downstream processing costs prevent its application on an industrial level. The other
side of the medal is that fossil fuels are still unbeatably cheap, which is not in the
sense of sustainability.
The most striking feature of algae are of course the nicely colored pigments.
They are gained by lipophilic extraction, e.g., n-hexane. The lipid fraction contains a
high variety of pigments such as chlorophylls, carotenoids, tocopherols, and sterols.
Chlorophyll amounts are usually about 0.5–1.5% of DW and can be used as food
and pharmaceutical colorants. Carotenoids in microalgae possess primarily photo-
protective (antioxidant) functions. Correspondingly, they are deployed as sources
of antioxidant activity in human and animal diets. Despite the high variety (>600
carotenoids known) in nature only a few were used commercially. For example, β-
carotene was successfully produced in open pond mass cultures of Dunaliella salina
already in the mid-1960s in the former USSR. The cell can accumulate up to 12% of the
product. Despite the current possibility of chemical synthesis of β-carotene (all-trans
isomers), there are still several β-carotene production plants from D. salina running
nowadays (e.g., Australia), because cis-isomers of ß-carotene can only be produced
naturally. Astaxanthin (AXT) (Figure 8.24) represents another carotenoid, which is
produced on a commercial scale as fish feed (accountable for red color e.g., of salmon
fillet) in Haematococcus pluvialis. In fact, it is the strongest natural antioxidant. In

Phycoerythrin

Phycocyanin

Allophycocyanin
Fig. 8.24: The specific antenna pigments of
PSII PSI cyanobacteria are the source of valuable pig-
ments.
 8.8 Products and processes | 215

some countries it is also available as food supplement (7,000 US$/kg!). Chemical syn-
thesis is possible, however with the debate of whether it is equivalent to biologically
produced material with different isomers.
Tocopherols depict another source of industrially interesting compounds and are also
found in the lipid fraction, e.g., α-tocopherol (vitamin E) has higher activity as an-
tioxidant if extracted from natural sources (plant or algae) than if chemically syn-
thesized (all-rac-α-tocopherol). The highest concentration of α-tocopherol is found in
photoautotrophically grown Euglena gracilis, so that this protist seems to be a promis-
ing source of naturally grown products. Moreover, other lipid soluble carotenoids such
as lutein, zeaxanthin, lycopene, and bixin are used in animal feeds, pharmaceuticals,
cosmetics, and food colorings, however to lesser extent.
Besides the lipophilic pigments, cyanobacteria and Rhodophyta also contain
water soluble fluorescent pigments called phycobiliproteins, where tetrapyrrole
chromophoric prosthetic groups, named phycobilins, are linked to a protein back-
bone. For commercial production of blue and red colorings phycobiliprotein (phy-
cocyanin, up 15% in cell mass) and phycobiliprotein (phycoerythrin) the cyanobac-
terium Arthrospira platensis (Spirulina) and the rhodophyte Porphyridium are most
commonly used, respectively. In fact, phycocyanin is the best blue color for food ap-
plications, e.g., in chewing gums, candies, dairy products, jellies, ice creams, soft
drinks, and others. It is also available as a blue colored drink for general health sup-
port. These pigments are also in smaller scale used as highly sensitive fluorescence
markers in clinical diagnosis and for antibody labeling. The high price makes pro-
duction in genetically engineered heterotrophs worth considering. A look at the light
harvesting complex of cyanobacteria (Figure 8.25) teaches us that the pigments are
well organized in staples. Overexpression of the strongly reductant phycobilins would
lead to disordered accumulation in the cytosol and therefore to cell damage. Note also
the close relationship (endosymbiosis) between cyanobacteria and Rhodophyta.
A further promising application for high value microalgae based products might
be represented by the synthesis of recombinant proteins for medical use. Although
mammalian cell culture (e.g., CHO) and bacterial expression systems currently dom-

O
H
3' OH
S
E E E E E
E E
E E

H 3
S
HO
O

Fig. 8.25: The pigment astaxanthin (3S, 3’S) as an example of a microalgae derived antioxidant. It
exists in different stereoisomers; asymmetric carbons are here marked as ‘E’ and ‘S’.
216 | 8 Microalgae – the solar cell factory

inate the market (with roughly US$100 billion/year) for therapeutic proteins, mi-
croalgae could represent an alternative promising platform. Microalgae possess all
the necessities for production of complex therapeutic proteins as other eukaryotes.
These include the presence of chaperones, disulfide isomerase required for protein
assembly, and last but not least the capability for posttranslational modifications.
Phototrophic production systems furthermore offer intrinsic security as no viruses
could infect the culture (and later the human applicant) and no other organic material
is in the medium, simplifying product separation. These qualities could make produc-
tion at lower cost possible, thus enabling the disposability of affordable quantities
of vaccines for developing countries. Despite the establishment of several successful
recombinant proteins (e.g., antibodies) in microalgae hosts, industrial production in
large scale and outdoor conditions (under natural illumination) has to be elucidated
in the near future.

8.8.3 Conversion of photon energy to chemical energy – producing fuels

by microalgae

Another group of lipids to be found in microalgae are neutral lipids (e.g., triacylglyc-
erol, TAG), which are the main storage compounds (besides starch). They accumu-
late in intracellular vesicles and can make up to 60% of DW. During the day the cells
accumulate these storage compounds in case photosynthesis is running faster than
growth processes. Nutrient deprivation can be a reason, which is actually often the
case in nature. During the night the cells can convert starch and oil into biomass pro-
vided all necessary minerals (N, P, etc.) can be taken up. Respiration on the storage
compounds and building of new cell compounds can lead to a decrease of 20% in dry
weight, which is not to be misunderstood as maintenance. To mimic these cycle in
technical processes microalgae can be cultivated under nitrogen limitation leading to
an excess of photosynthetic energy and consecutively to lipid accumulation. A related
experiment is shown in Figure 8.26.
TAGs are interesting as future solar fuels, however comparably high cultivation,
harvesting, and downstream processing costs prevents its application on an industrial
level. The other side of the medal is that fossil fuels are still unbeatably cheap, which
is not in the sense of sustainability.
In analogy to the yield coefficient for the carbon source in the heterotrophic
case, an energetic efficiency ηComp,hν is defined, which is the ratio of energy stored in
biomass or products and light energy received. These efficiencies are commonly called
photoconversion efficiency (PCE [MJ/MJ]), requiring an additional index of definition
of the system boundaries. The stored energy in chemical compounds can be given as
heat of combustion HC0 [MJ/kg] (higher heating value). The values for microalgae have
to be measured in principle for each species under investigation and specifically for
the macromolecular composition. Educated guesses for the compounds are 17 MJ/kg
 8.8 Products and processes | 217

200 1.0 4 7

cStarch [g/L]
0.8
Itrans Em-2s-1

150
5
NO3 [g/L]

cX [g/L]
0.6
100 4
2

cLipid [g/L],
3
0.4
50
2

0.2
1
0

0.0 0 0
0 3 6 9 12 15 18 21 24 27
Culvaon me tc [d]

Fig. 8.26: Batch experiment under nitrogen deprivation; even after growth stops lipid accumulates
constantly without any remarkable loss of photosynthetic activity. The slope given as lipid produc-
tion over absorbed photons corresponds to a PCE of 4%.

for carbohydrates and proteins, and 30 MJ/kg for lipids, so a mean value of a cell could
be 25 MJ/MJ. Check these values on the label of food packages in the supermarket. With
this definition we can calculate expected harvests under given light conditions.
Now PCEComp,hν can be calculated (8.20) for a given time interval considering the
amount of biomass or product produced during this interval and the light energy re-
ceived as integration of the spectral light intensity If over the frequency range of in-
terest:
∆ ⋅ HC,Comp
0
⋅ mComp
PCEComp,hν = ∆t f (8.20)
∫0 ∫f hν,max h ⋅ If (f hν ) ⋅ df hν ⋅ dt
min

Important to note is that this value strongly depends on the system boundaries cho-
sen. Here we start with the smallest subsystem, the conversion of light into starch as
the primary product. In the biological context PCEStarch,hν is called photosynthetic ef-
ficiency (PE). The theoretical maximum can be estimated from the stoichiometry of
photosynthesis (8.21). To produce one molecule of glucose six CO2 have to be fixed,
each needing at least eight photons, as measurements revealed. In best the case, these
photons are divided between wavelengths λ1 = 680 nm (PSII) and λ2 = 700 nm (PSI)
contributing E hν = NA ⋅h⋅c⋅λ−1 = 6.02⋅1023 mol−1 ⋅6.63⋅10−34 J ⋅ s⋅299.8⋅106 m ⋅ s−1 ⋅
1/680 nm−1 = 176 kJ/mol and 171 kJ/mol respectively:
kJ
2805 mol ⋅ 1 mol
PEtheor = kJ kJ
= 0.34 (8.21)
6 ⋅ 4 ⋅ mol ⋅ 176 ⋅ mol + 6 ⋅ 4 ⋅ mol ⋅ 171 ⋅ mol
218 | 8 Microalgae – the solar cell factory

This is actually better than a PV device. However, in practice more photons are needed
(10–12 per mole CO2 ) reducing the value to about PE = 0.24 MJ/MJ. Furthermore, only
photons are considered really active and monochromatic in the red range. By enlarg-
ing the balanced system in the direction of a larger spectrum, e.g., the spectrum of sun-
light, efficiency is reduced. All photons have in principle the same quantum yield in
photosynthesis. That means that e.g., ‘blue’ photons being of higher energy than ‘red’
ones contribute a lower energetic efficiency. Excess energy is dissipated via the already
mentioned nonphotochemical quenching NPQ. Multiplying PE by 0.75 to account for
the average of photon energy yields PCEstarch,PAR = 0.18 MJ/MJ. These value are the
highest measured values under ideal lab conditions. Further enlarging the calcula-
tion basis by including the IR part of sunlight yields PCEstarch,sun = 0.1 MJ/MJ. Now we
consider the whole cell as a system. Photons with wavelengths between the maximum
absorption wavelength transfer their energy firstly e.g., to carotenoids, from where it
is further transferred to the reaction center. This process is less effective thus reduc-
ing overall efficiency. Other photons may be absorbed by other cell components. The
main energy loss on the cell level includes the synthesis of cell material from starch
or lipids, which happens similarly as in heterotrophic microorganism by respiration
on one part of the storage compounds and using the gained energy for growth with
the leftover part of starch as a carbon source. As each ATP dependent metabolic step
has an energetic efficiency < 100%, which reduces the overall efficiency of biomass
produced by sunlight absorbed to PCEX,sun = 0.07 MJ/MJ. With the next step in en-
larging the system boundaries we look at the suspension, where light saturation and
dark zones exists. In outdoor applications reflection, loss on ground, and other ef-
fects also decrease the practical efficiency to PCE = 0.05 MJ/MJ, a commonly accepted
maximum value for microalgal energy production based on sunlight (including IR) to
biomass. Note that a productivity of 100 t/ha/year will decrease to 80 t/ha/year when
oil production with higher heat of combustion is induced. This is without loss in en-
ergetic efficiency. Unfortunately, the effect of decreasing yield while enlarging system
boundaries is quite common in process engineering. Think e.g., of the efficiency of
pumps and compressors not considered in our theoretical argumentation. At least we
could confirm that microalgae convert light energy to chemical energy remarkably bet-
ter than terrestrial plants, e.g., sugar cane with PCE = 1–2%.

8.9 Outlook

Further development has to go to lower costs for reactors and auxiliary energy to
serve markets with increasing volume. The high value products are nutritional sup-
plements, feed for aquaculture, food, bulk chemicals, and fuels. The last item is the
most difficult one as costs are low and energy has to come to a minimum. New re-
actor developments and better integration into the environment hold the potential
for us to proceed along these stepping stones. With respect to strains more process
 8.10 Questions and suggestions | 219

oriented development should take place. One issue is reducing antenna pigments
to minimize mutual shading and NPQ or simply to prevent flocculation. This way of
thinking is called ‘domestication’, meaning going through a development analogous
to that of crops during the last 10,000 years but here in a few decades. Biofilm reactors
and extracellular products are desired as well. Besides easier harvesting this could
be operated at higher PCE of e.g., 10%, as growth can be set to zero. Microalgae as
aquatic biomass will severely contribute to the transformation of a petrochemical to
a biobased economy.

8.10 Questions and suggestions

1. Think about the physiological interaction between light and human senses:
Why does a green laser pointer look brighter than a red one? Why does sun-
light feel warmer than LED light of same intensity?
2. Set up a calculation table for a generic photobioreactor and an environment
to calculate anticipated production yields for different biomasses.

1. The human retina is more sensitive to green light, and the power of the laser
pointer is actually limited by regulation to 1 W. The IR part of sunlight (called ther-
mal radiation) is not warm as such, but can penetrate the upper layer of the skin
and is absorbed in the region of heat receptors, while other colors are better re-
flected.
2. See supplementary material.
9 Continuously operating bioprocesses – production
under steady state conditions
From the exercise in Chapter 7 on the fed-batch process it became clear that highest
productivity can be reached by miniharvesting only a small amount out of the reactor
and fill up with fresh medium. The same holds for batch processes. The idea is now to
go to infinitesimally small harvesting amounts but very frequently taken. Going to the
mathematical limit finally leads to simultaneous feeding and harvesting with the same
rate to keep the volume constant. The reactor configuration, as shown in Figure 9.1,
gives the basic variables. Assuming an ideally mixed reactor, the concentrations inside
the reaction volume are of course the same as in the outlet. As there is no defined time
of harvesting and the culture is active for a long time, this type of process is called
continuous process or also (but not strictly correctly) continuous cultivation.

Control volume

System boundary

Fig. 9.1: Reactor configuration for a continuous bioprocesses; usually the outlet tube is mounted in
such a way to suck surplus medium from the surface.

In this chapter firstly the balance equations are derived and discussed. This is done as-
suming steady state conditions, which is one of the features of continuous processes.
Another feature is that large continuously available amounts of medium can be han-
dled. That is the case in wastewater treatment but also other large scale processes.
Consequently, most of the processes are from these areas of application. Some struc-
tural modifications are possible for further process intensification. In chemical engi-
neering most of the processes are continuously operated. This is basically not the case
in bioprocess engineering despite all potential advantages. At the end of the chapter
some aspects of the current debate about the reasons for this observation are reflected.
Thermodynamically speaking the continuously operated reactor is an ‘open’ sys-
tem, receiving energy and material from outside and giving used material with typi-
cally less energy back to the environment. Thereby, the system can keep a state of low
entropy at the cost of the environment. In this general sense lakes, ecosystems, but
also each living entity can be regarded somehow as a continuous process.

https://doi.org/10.1515/9783110315394-009
 9.1 Setting up stationary balance equations – understanding process behavior | 221

9.1 Setting up stationary balance equations – a good start in

understanding process behavior

To get an adequate description of the process we draw a balance around the system
volume counting the amount of all compounds going in or out through the system
boundary. For dynamic reaction systems a general scheme for material balances is:

Accumulation = input − output +/− reaction.

For the single compounds substrate and biomass the material mass balances read:
dmS (t)
= qS,f (t) ⋅ cS,f − qS,f (t) ⋅ cS (t) − rS ⋅ m X (t) (9.1)
dt
dm X (t)
= −qS,f (t) ⋅ m X (t) + r X ⋅ m X (t) (9.2)
dt
The empty space in these equations is a hint to the missing input. Like in a bank ac-
count, we do not have to pay money in, but can make a withdrawal and get an increas-
ing current balance, as long as the account status is high and the interest rates are as
well, here the concentration and growth rate of the microorganisms.
While masses are conserved quantities, we are more interested in concentrations.
As the working volume VR is constant, it is possible to divide by it, getting:
dcS (t)
= D (t) ⋅ (cS,f − cS (t)) − rS ⋅ c X (t) (9.3)
dt
dc X (t)
= −D (t) ⋅ c X (t) + r X ⋅ c X (t) (9.4)
dt

The ‘dilution rate’ D [h−1 ] is the defined as q/V R . In the case where nothing reacts in the reactor,
an indicator substance inside would be transported out and ‘diluted’ by freshly fed medium. The ob-
servable concentration would decline exponentially following the time constant τ = 1/D, called mean
residence time.

The primary intention of dealing with the continuous process as such was to hope
for steady state conditions. So for now it is assumed that the process is in indeed in
steady state and the concentrations do not change over time. This will of course only
be the case if D is constant for a long time as well. Furthermore, we are interested in
the dependency of substrate and biomass concentrations on the dilution rate.
!
D ⋅ (cS,f − cS ) − rS (cS ) ⋅ c X = 0 (9.5)
!
r X (cS ) ⋅ c X − D ⋅ c X = 0 (9.6)
This is a set of two nonlinear (product of variables) algebraic equations for the two
unknown variables c X and cS . The exclamation mark reminds us that the equality is
not given explicitly following a mathematical deduction of the left side of the equa-
222 | 9 Continuously operating bioprocesses – production under steady state conditions

tion, but that it is for the moment only our own demand and we have to check under
which conditions it will be reality. A closer inspection or explicitly solving the equa-
tions brings up unexpected results:
Looking at the stationary biomass balance Equation (9.7) it can be noticed that it has
two formal solutions, namely:

c X = 0 and r X (cS ) − D = 0 (9.7)

Putting the first solutions into the substrate balance we get the corresponding value
for the substrate concentration:
cS = cS,f (9.8)
Firstly, we try to understand this trivial solution c X = 0, cS = cS,f . This case is in-
deed not only mathematically possible but happens in reality if we forget to inocu-
late the reactor or if the dilution rate exceeds the maximum possible specific growth
rate. Biomass grows more slowly than it is withdrawn from the reactor and vanishes
completely after some time. The situation is therefore called the ‘washout case’. From
where does ‘the reactor know’ which one of the two solutions is true? This depends
on the starting conditions, a piece of information that got lost in the moment, when
we set the derivatives to zero. So for application the history of the cultivation has to be
checked.
Now we come to the second and obviously more relevant and interesting opera-
tion solution. Most important in understanding continuous bioprocesses is that the
specific growth rate μ = r X = D. So by changing the dilution rate, we can directly ma-
nipulate the specific growth rate. This a unique feature, as we can keep μ for a very
long time on values of interest. For biological investigations this is great as adaptation
is completed and there are only minor physiological changes. Now of course the mi-
croorganisms have a say. How can ‘the microorganism know’ that they have to keep
μ at D? Substrate uptake rate and specific growth rate depend of course on substrate
concentration. So a minimum requisite to maintain a constant μ is that the substrate
concentration stabilizes on a level allowing the cells a specific substrate rate in accor-
dance to the required μ. For further calculations we need to specify a kinetics as a link
between substrate and biomass, for now we take the standard set of equations (4.43)
and (4.44).
Without maintenance, the second solution reads:
cS D ⋅ kS
r X (cS ) − D = 0 → y X,S ⋅ rS,max ⋅ − D = 0 → cS (D) = (9.9)
cS + kS y X,S ⋅ rS,max − D

This equation looks different for different kinetics.

From the substrate balance we finally get:

c X (D) = y X,S ⋅ (cS,f − cS (D)) (9.10)

The solutions are plotted in Figure 9.2.

 9.1 Setting up stationary balance equations – understanding process behavior | 223

100
cS [g/L]

80

60
cx [g/L]

40

20

0.0 0.1 0.2 0.3 0.4 0.5 0.6

D [1/h]

Fig. 9.2: Stationary states of glucose and biomass concentration over dilution rate. rS,max =
1 g ⋅ g−1 ⋅ h−1 ; kS = 1 g ⋅ L−1 ; y X,S = 0.5 g ⋅ g−1 ; c S,f = 1 g ⋅ L−1 .

Indeed, substrate concentration increases with increasing D. The higher D already is,
the higher the necessary increase of cS . What is actually seen is the Michaelis–Menten
kinetics for substrate uptake, but here not as rS = f(cS ) but the inverse function cS =
f(D = μ). So this is the important ‘second’ solution, assuming Michaelis–Menten for
substrate uptake and Pirt equation without maintenance. Turn the plot and hold a
mirror vertically on the plot, then you will recognize the kinetics!
One detail deserves attention: when the dilution rate D approaches the theoreti-
cal maximum specific growth rate μmax = y X,S ⋅ rS,max the denominator approaches
zero and the necessary substrate concentration infinity. As the maximum possible
substrate concentration of cS,f , μ max can be reached only approximately we get the
washout case already for D < μmax .
For low substrate concentrations and dilution rates the biomass concentration is
nearly constant close to the maximum possible value y X,S ⋅ cS,f . For higher dilution
rates a higher substrate concentration is necessary (as explained above). This sub-
strate is no longer available for growth and the biomass concentration drops.
The volumetric productivity for biomass is generally defined as the produced
amount of biomass per volume and time:
∆t
∫0 q ⋅ c X ⋅ dt q ⋅ c X ⋅ ∆t
PV,X = = = D ⋅ cX (9.11)
VR ⋅ ∆t VR ⋅ ∆t
The volumetric productivity is high for high dilution rates and for high biomass con-
centrations. But for now we have to accept that both values do not reach their respec-
tive maximum at the same working point.
224 | 9 Continuously operating bioprocesses – production under steady state conditions

40

35

30 operang points
RS

25

20
TS

15

10

0
0 20 40 60 80 100

cS [g/L]

Fig. 9.3: Volumetric transport TS [g ⋅ L−1 ⋅ h−1 ] and reaction rates R S [g ⋅ L−1 ⋅ h−1 ] over substrate
concentration for a process with substrate inhibition kinetics, parameters as above, D = 0.3 h−1 .

Like most bioprocesses the continuous process is a transport/reaction system. The

transport term for substrate TS = D ⋅ (cS,f − cS ) includes an inflow and an outflow
component. The reaction term RS = rS ⋅ c X embraces the biomass concentration and
the biomass activity as a physiological component. Both terms (rS including substrate
inhibition) are plotted in Figure 9.3 versus a virtually given substrate concentration.
Possible stationary operating conditions require that the same amount of sub-
strate is transported as used up by the cells, hence TS = TR , given by the points of in-
tersection of the two curves. Firstly, we look at the left working point for low substrate
concentrations. Now we imagine that substrate concentration is lower than the con-
centration necessary to maintain the wanted operation point. The cells will grow with
μ < D, and biomass concentration slowly drops below the precalculated value. This
leads to decreased substrate consumption, and the transport term is higher than the
reaction term leading to an increase of the substrate concentration. Finally the system
will again approach the point of equilibrium. The same happens accordingly for sub-
strate concentrations arbitrarily higher than the value of equilibrium. This qualitative
stability analysis shows that the culture, even when disturbed and not at equilibrium,
will autonomously find its way back to steady state without further control action from
our side. This answers the question of what makes the substrate concentration stay at
the required value. Such a stable operation is called a ‘chemostat’ as all compounds
involved are in steady state. A graphical representation of this feedback loop is shown
in the simulation example. The washout case at high substrate concentrations is sta-
ble as well. In cases of substrate inhibition a third theoretical working point exists,
which is unstable.
 9.1 Setting up stationary balance equations – understanding process behavior | 225

D= 0.4 D= 0.5
25 50

dcx/dt [g/(L.s)]
10
20

cx [g/L]
cS [g/L]

5
15 45

dcS/dt [g/(L.s)]
10
0

5 40
-5

0
15 20 25
Time [s]

Fig. 9.4: Dynamic simulation with the same parameters as Figure 9.2. The derivatives are a measure
of how close the system is to equilibrium.

But how long does it take to reach the equilibrium especially after the initial inocula-
tion or after changing the dilution rate? To answer this question a dynamic simulation
helps, which is shown in Figure 9.4.
The initial starting conditions after the inoculum is really close to steady state. Fill-
ing the reactor in the beginning with full medium would be counterproductive. Only
in cases where not enough fresh culture for inoculation is available can the continu-
ous process be preceded by an initial batch (D = 0). Luckily, for most of the dilution
rates steady state conditions are really reached, which is not self-evident looking only
at the stationary balance equations. A rule of thumb says that waiting time is around
five residence times 1/D. That is fortunately not true as can be seen from the deriva-
tives, but the real time constants have to be calculated, see below. One has to be careful
as adaption could last for several cell generations. Further we observe that the time to
reach steady state increases with higher dilution rates. To understand this behavior
we have to linearize the system equations for different substrate concentrations and
extract the dominant time constant, for details see the calculation in the supporting
material. In fact, transport and reaction contribute to transitional system dynamics. At
low dilution rates and substrate concentrations reaction dominates and leads to fast
time constants (τ ≈ −rS,max /k S ⋅ y X,S ⋅ cS,f ), while at high dilution rates and substrate
concentrations the hydrodynamic time constant τ = 1/D dominates. In the special
case of D = μ(cS,f ) growth is very insensitive to substrate concentration and the time
constant of biomass goes to infinity. It can indeed be tormenting to wait to see whether
a culture is stabilizing or slowly being washed out.
226 | 9 Continuously operating bioprocesses – production under steady state conditions

9.2 Ethanol production in a continuous process –

the window of operation

As a first data example we investigate ethanol production. It is often carried out in

batch processes as we have already seen. As biofuels are produced in large amounts,
the process is not too complicated, and downstream may be carried out continuously
as well, it seems to be a good idea to apply continuous cultivation. An X/D diagram for
ethanol production with the bacterium Zymomonas mobilis is shown in Figure 9.5. This
bacterium, known from ‘cider sickness’, has been proposed as an alternative ethanol
producer. Although yield and productivity were very good, the attempts failed because
the bacterium could not use fructose properly and grow in a neutral pH range abetting
contamination with other bacteria. This is an example of how secondary aspects can
also be a reason to prevent a ‘go’ for new processes. Here it is a good example to discuss
chemostat cultivation.
The experiments were carried out with cS,f = 100 g/L. So the biomass concentra-
tion reaches about 3 g/L, which corresponds to a yield of 0.03 g ⋅ g−1 . The maximum
ethanol concentration is around cP = 50 g/L, what also fits to a typical anaerobic fer-
mentation scheme. The data designated as CTR/D is a virtual variable assuming that
the produced CO2 would behave like a liquid product. This is a trick to visualize the
relation of the two products ethanol and CO2 . Both products are formed in the stoi-
chiometric relation 1:1, so we expect a mass relation of MGEth /MGCO2 = 46/44. This is

5 100
Ethanol cp [g/L]

14
90

4 80 12

70
10
3 60
Pv,Eth [g/(Lh)]
CTRD [g/L]

8
Biomass cx [g/L]

50

2 40 6

30
4
Glucose cS [g/L]

1 20
2
10

0 0 0
0.0 0.1 0.2 0.3 0.4 0.5

D [1/h]

Fig. 9.5: X/D diagram of a continuous cultivation of Zymomonas mobilis. Besides biomass, sub-
strate, and product, the volumetric productivity is also shown.
 9.2 Ethanol production in a continuous process – the window of operation | 227

100
cS [g/L]

80

60
cx [g/L],

40

20

0.0 0.1 0.2 0.3 0.4 0.5 0.6

D [1/h]

Fig. 9.6: Simulation of a chemostat with microorganisms exhibiting remarkable maintenance μ m =

0.1 h−1 , other parameters as above.

approximately true; maybe a part of the ethanol evaporates together with the CO2 into
the off-gas. But a further inspection of the data shows several differences to the stan-
dard curve we deduced in the last paragraphs. At low dilution rates c X is not constant
but decreases for decreasing D. Obviously, the cells take up glucose but do not form
much biomass from it. This reminds us of the maintenance term in the Pirt equation.
A simulation (Figure 9.6) shows that a maintenance term indeed leads to this typical
curve.
The second salience is the slow decrease of c X for increasing D at the range of
higher dilution rates. In the standard simulation this drop of biomass concentration
was sharper. It is indeed not plausible that an increase of cS from 10 g/L to 50 g/L
helps the cells to increase the specific growth rate from 0.25 h−1 to 0.35 h−1 . At such
high substrate concentrations we can exclude substrate limitation. Also cP drops in
this range. The reason for this observed behavior of slow biomass decrease has been
identified as diminishing ethanol inhibition. The volumetric productivity PV,P = D ⋅ cP
decreases as well. As a ‘lesson learned’ it can be stated that in principle ethanol can
be produced in a continuous process. But for inhibiting products high product con-
centrations and high volumetric productivities cannot be reached simultaneously.
The optimum working point would be somewhere between the highest produc-
tivity and the highest product concentration depending on reactor costs and costs for
rectification. This leads to lower limits for both ethanol concentration and productiv-
ity. Other limits are given by the sugar concentration in the available substrate, or the
total amount to be produced in a given time ordered by a customer. These limits form
a ‘frame’ of sensible working points visualized as an ‘operating window’. In aerobic
processes this could be for example the highest possible oxygen transfer rate (OTR), or
228 | 9 Continuously operating bioprocesses – production under steady state conditions

the highest possible biomass concentration (e.g., 100 g/L). Also the shift of the intra-
cellular product concentration to a specific point could be a limit if it is possible only
at the cost of overall productivity. In other industrial areas the integrity operating win-
dow (IOW) is a set of limits used to determine different variables that could affect the
integrity and reliability of a process unit. This standard was set up by the American
Petroleum Institute.
What could be a way out of such a window? As long as no changes in the mi-
crobial kinetics are possible, e.g., by strain development, structural changes in the
process have to be foreseen. One strategy could be to decompose otherwise closely
interlocked parts of the system like stoichiometry or other kinetically coupled physi-
ological states. An example was the fed-batch process where different phases of the
process – growth and production phase – were decoupled. For the current example of
continuous ethanol production not a timely but a spatially decoupled process struc-
ture could be envisaged. This means in practice the employment of a second reactor
as shown in Figure 9.7.
In the first continuously operated reactor optimum growth conditions are ad-
justed. This means low or medium ethanol and high substrate concentrations. Biomass
and unused substrate go into the second reactor, where additional substrate could
be fed. Here high product concentrations could be achieved at low growth rates. As
biomass comes in from the first reactor μ = D does not hold and the dilution rate can
be higher than the specific growth rate. Even zero growth could be an option, where
the ethanol concentration may reach values inhibiting growth completely. Substrate
turnover is ensured only by maintenance. This can even be an advantage as more
carbon is allotted to the product.

Fig. 9.7: Two reactors coupled in series to increase ethanol concentration and productivity.
 9.3 Enzymatic processes – a simple example for continuous bioprocess operation | 229

9.3 Enzymatic processes – a simple but effective example

for continuous bioprocess operation

Directly from the equations above we see that one disadvantage of continuous pro-
cesses is the permanent discharge of biomass. Especially in cases where product for-
mation is at the cost of metabolic energy and therefore low specific growth rates, this
is not acceptable. Extracellularly produced enzymes are an example. Here the specific
advantage of easy harvesting during the process would additionally be lost. In cases
where resting cells or enzymes are used as biocatalysts continuous production in the
form described would be impossible. A way forward is changing the system to keep
the catalysts inside reactor independently from medium and hydrodynamic retention
time. The resulting reactor structure is given in Figure 9.8. Such a measure is called
either cell recycle, in cases of an external separation step, or cell retention, in cases
where the cells are kept inside the reactor by immobilization or by submersed mem-
branes. Basically, we have now established a perfusion system. This is well known
from chemical engineering, where the catalyst has to be kept in the reactor in any
case.

qb

qout

qf qp
qf,S
cS,f
qr cX,r

Fig. 9.8: Structure of a bioreactor with cell recycle or cell retention; q b = bleeding flow; qf filter feed
flow; q p cell free permeate flow; q r retentate flow.

A flow of fresh medium (qS,f ) enters the reactor, where the substrate is used by the
growing cells. The reactor suspension is discharged via a solid-liquid separation step,
where the ‘solids’, namely the cells, are fed back into the reactor in highly concen-
trated flow q r . The permeate flow q p contains mainly the water from the input and
a possible product. A ‘bleeding’ stream has to be foreseen to simulate an unavoid-
able loss of cells or to discharge cells in excess. The total flow out qout = qS,f = q.
Now we can set up the mass balance equations around the control volumes ‘reactor’
(dynamic) and ‘filter’ (static). The complete deduction is given in the supplementary
material, here we select only two relationships.
230 | 9 Continuously operating bioprocesses – production under steady state conditions

Firstly, a recycle ratio R is defined:

qp qp
R= = (9.12)
qout q b + q p
R = 0 means no recycle, R = 1 means that there is absolutely no biomass in the
effluent.
For biomass we then get:
dc X (t)
= r X ⋅ c X − D ⋅ (1 − R) = 0 → r X = D ⋅ (1 − R) (9.13)
dt
A plausibility check for R = 0 leads to μ = D as in the regular continuous process, for
R = 1 the result is μ = 0 and the biomass goes theoretically to infinity. In practice,
recycle ratios around 0.9 are commonly applied. In cases where product formation is
directly coupled to growth, R should be even smaller.
Organisms and enzymes have the unique ability to perform enantiospecific reac-
tions, which are not easily performed in chemical synthesis. Stereospecificity is indis-
pensable in biochemical production, as only the L- or D-form of a substance is biolog-
ically active, while the other one is in the worst case toxic. Amino acids have at least
one stereocenter and are a good example for a closer look.
Examples of continuous production processes employing immobilized enzymes
or resting bacteria include the production of amino acids. L-aspartic acid can be pro-
duced by enantioselective addition of ammonia to fumaric acid, a substance easily
produced chemically at low cost. Racemic amino-aminocaprolactam (ACL) can be hy-
drolyzed to give L-lysine by an immobilized L-ACL-hydrase; R-ACL is racemized to re-
place S-ACL by a racemase until the chemical precursor is completely used up. A gen-
eral scheme of such a process is shown in Figure 9.9.
Racemic mixtures of amino acids could in principle be produced chemically. The
main problem is the separation of the two forms. Here the stereospecificity of enzymes

Chemical
synthesis
Filtraon Heat exchanger
Acetyl-DL-
amino acid

Racemizaon
Separator Crystallizaon Vaporizer
Column reactor
with immobilized
aminoacylase
L-Amino acid

Fig. 9.9: Flow sheet to obtain L-amino acids from a D-L-precursor, here with deacetylation and crys-
tallization as separation step.
 9.4 Biogas production via anaerobic digestion | 231

comes into action. The idea is to produce a precursor chemically and perform only
the last step enzymatically. The L-amino acid and the remaining D-precursor can be
separated e.g., by crystallization as they are two substances with different physical
properties. This approach is so far an example of integration on the process level, as
a bioprocess is coupled to a chemical process. Enzyme reactors can also be handled
in an industrial environment, where no know-how or competence of microbial culti-
vation is available.

9.4 Biogas production via anaerobic digestion

Fermentative biogas generation via anaerobic digestion (AD) is a naturally occurring

process that is readily observed when organic matter decomposes in anoxic milieu,
e.g., in natural wetlands and rice fields, as well as the intestinal tract of ruminants and
termites. Human beings have been using anaerobic digestion processes for centuries,
however the first documented digestion plant was constructed in Bombay, India in
1859. The first usage of biogas from a digester plant for street lighting was reported
in 1895 in Exeter, England. Nowadays, such processes are regarded not only as tech-
niques for treatment of sewage biosolids, livestock manure, and concentrated wastes
from the food industry, but also as a potentially significant source of renewable fuel.
Due to the large waste streams AD is carried out as batch, where the substrate is par-
tially solid.
Organic matter is usually composed of complex polymeric macromolecules (of-
ten in particulate or colloidal form), such as proteins, carbohydrates, and lipids. The
anaerobic digestion process converts organic matter to the final products (methane
and carbon dioxide), new biomass, and inorganic residue. Several groups of microor-
ganisms (anaerobic bacteria and archaea) are involved in organic substrate transfor-
mation to CH4 , CO2 , and H2 O, and the overall process comprises multiple conversion
stages with many intermediate products. Commonly, the conversion is simplified to
four successive steps, running in parallel in the process: (I) hydrolysis; (II) fermenta-
tion or acidogenesis; (III) acetogenesis; and (IV) methanogenesis (Figure 9.10).
Biological biogas formation is a sequential process in the sense of consecutive
biochemical steps and microbial consortia involved. During anaerobic digestion in
biogas plants (Figure 9.11), organic biopolymers such as lipids, polysaccharides, and
proteins are converted by anaerobic hydrolytic bacteria into less complex compounds,
which can then be further used by other microorganisms. According to microbiologi-
cal examinations, hydrolytic species are found in a broad range of bacterial phyla and
many of these bacteria have developed cell bound multienzyme complexes, known as
the cellulosomes, for the decomposition of cellulose- and hemicellulose-containing
substrates. Many bacterial species do not have cellulosomes, however they are able to
secrete free hydrolases, containing multiple catalytic domains or they produce many
other enzymes such as glucanases, hemicellulases, xylanases, amylases, lipases, and
232 | 9 Continuously operating bioprocesses – production under steady state conditions

Biopolymers

Hydrolysis
Fay acids, amino acids, mono- &
oligosaccharides

Acidogenesis
Short-chain organic acids
& alcohols

Acetogenesis
Acetate
CH3COOH CO2 H2

Acetoclasc Hydrogenotropic
Methanogenesis Methanogenesis Fig. 9.10: Flow diagram of the anaerobic diges-
CH3COOH CH4+ CO2 CO2 + 4 H2 CH4+ H2O tion process of organic matter; ellipsoids indicate
substances with lighter color meaning smaller
Biogas (volatile) molecules; rectangles are summarized
(CH4, CO2) conversion steps.

Fig. 9.11: Arial view of a biogas plant in North Germany, maize silage clamps on right upper side and
digester/fermenter tanks in the center (© Martina Nolte).

proteases for efficient biomass hydrolysis. The abundance of every hydrolytic bacterial
species is dependent on the inoculum type of the digester and substrate applied, thus
in biogas plants the members of the phyla Firmicutes and Bacteroidetes are the most
commonly found, while others belonging to Fibrobacteres, Spirochaetes or Thermo-
togae are less abundant. Thereby the members of the genus Chlostridium (Firmicutes)
are described as usually dominating the bacterial community in the biogas plant.
In the next steps, acidogenesis, fermentative bacteria convert the breakdown
products of hydrolysis to simple carbonic acids (e.g., propionate, butyrate, acetate,
formate, succinate, and lactate), alcohols (e.g., ethanol, propanol, and butanol), and
other compounds (e.g., H2 , CO2 , VFAs, and ketones). Some of these products (e.g.,
 9.4 Biogas production via anaerobic digestion | 233

fatty acids longer than two carbon atoms, alcohols longer than one carbon atom,
and aromatic fatty acids) are then used by acetogenic or syntrophic bacteria within
the acetogenesis step for the conversion into acetate and C-1 compounds. Hydrogen-
producing bacteria, like the homoacetogenic bacteria Acetobacterium woodii and
Clostridium aceticum are usually described as performing the acetogenesis, resulting
in the generation of acetate, CO2 , and H2 . The final step, methanogenesis, is per-
formed exclusively by methanogenic archaea, whereby two functional groups are
involved; namely acetoclastic (utilizing acetate) or hydrogenotrophic (utilizing H2 ,
CO2 , or formate). Only few species are acetoclastic methanogens and thus able to de-
grade acetate into CH4 and CO2 , belonging to the order Methanosarcinales (e.g.,
Methanosarcina barkeri and Methanotrix soehngenii) and Methanococcales (e.g.,
Methanonococcus mazei), whereas all methanogenic archaea are able to use hydrogen
to form methane. Methanogens of the orders Methanosarcinales, Methanomicrobiales
and Methanobacteriales are usually the most abundant within the archaeal subcom-
munity.
Overall digestion speed depends on the interplay of all involved microorganisms;
the slowest step (often hydrolytic phase) is often also the rate limiting step for the
entire process. Despite the complexity of anaerobic fermentation by involvement of
different microorganisms and degradation steps the process is surprisingly frequently
successfully used for diverse applications (methane production from biomass, water
remediation, etc.). The main reason for this fact might be the age of the process (few
billions years), and correspondingly concurrent evolution of the interplay of involved
microorganisms and their metabolic self-regulation mechanisms. The optimal micro-
bial community for maximal efficient degradation (under these anaerobic conditions)
is permanently selected in these for the most part open (to the environment) systems.
Never the less, this process possesses like others some limitations, which have to be
taken into account for each application. An efficient AD process demands that both
substrate degradation and methanogenesis are balanced. This can be accomplished
by slow adaptation of the community to the desired substrate (stepwise adaptation
duration of half a year to a full year). This is sometimes not respected for different
reasons and the fermentation process collapses.
In the following some scenarios will be discussed that can lead to process im-
balances. Foremost, if the first process steps hydrolysis, acidogenesis, and/or aceto-
genesis are too fast, the fermentation can fail due to intermediate product inhibition
(scenario I). This can happen by application of easily degradable substrates (e.g., glu-
cose) since proliferation rates of methanogenic archaea are slower than those of the
other bacteria in this process. The volatile fatty acid (VFA) concentration (intermedi-
ate product) rises within the digester, and the pH drops below the optimal range (pH
6.5 to 8.5) for methanogenic archaea, which can lead to decreased methane forma-
tion rates and subsequently further accumulation of VFA, which lowers the pH until
the process is completely inhibited. If the methanogenesis runs too fast, methane pro-
duction is limited by the hydrolytic stage, however the process remains stable. Thus,
234 | 9 Continuously operating bioprocesses – production under steady state conditions

the rate limiting step depends heavily on the particular substrate used for biogas pro-
duction. Another inhibition scenario occurs when there is an imbalance of nitrogen
to carbon ratio (C/N ratio) of the substrate. An ideal substrate for anaerobic fermen-
tation should have a C/N ratio in the range of 15 to 30. When the C/N ratio is higher
than 30 microbial growth can be limited and the substrate is not digested completely
(scenario II). If C/N ratio is lower than 15 (scenario III), the concentration of ammo-
nia could increase during the continuous fermentation process over inhibitory levels
of ~ 1700 mg/L total ammonia nitrogen (TAN), where it can have an inhibitory effect,
especially on methanogenic archaea. In fact, the inhibitory effect is caused by free
ammonia nitrogen (FAN), which is part of TAN alongside with NH+4 according to the
dissociation constant dependent on pH and T. The amount of FAN (NH3 -N) is depen-
dent on TAN concentration, pH, and temperature (9.14) according to NH3 – NH+4 + OH−
equilibrium:
TAN ⋅ 10pH
FAN = 63.44 (9.14)
e 273.15+T + 10pH
The temperature is usually regulated to a constant level (mesophilic 35–45 °C or ther-
mophilic 45–60 °C) within the digester, so that this factor is not relevant for this sce-
nario. However, the pH is influenced by the NH4 -N concentration, since its accumula-
tion can strongly increase the pH in low buffered solution. Higher pH in turn leads
to higher dissociation of ammonium (NH4 ) to highly toxic ammonia (NH3 ), which
inhibits microorganisms within the digester (first of all, the methanogens) already
in low concentrations (50 mg NH3 -N/L). Interestingly, the inhibition of methanogen-
esis leads to the accumulation of intermediate products (VFA) as described above
(scenario I), and the pH drops alongside decreasing the concentration of highly toxic
FAN. Unfortunately, this self-regulation mechanism is very inert and in some cases
the inhibitory potential of individual factors (TAN osmotic stress, VFA acidification)
are combined and result in a complete inhibition of the fermentation process (at pH
below 6.4). Despite this biochemical self-regulation, ammonia inhibition is one of the
common reasons for misbalance in industrial applications, since the adaption time
for this more indolent biological system is not always applied in a sufficient manner.
Other curtail process management parameters for an efficient operating AD pro-
cess are represented by organic loading rate OLR [kg ⋅ m−3 ⋅ d−1 ], hydraulic retention
time HRT [d−1 ], and solids retention time SRT [d−1 ], whereby in common biogas plants
operated as continuous stirred tank reactor (CSTR) HRT is equal to SRT since no sub-
strate/water separation is carried out. For instance, the rapid increase of ORL, es-
pecially of easily digestible substrate, would cause fast acid formation and pH drop
like the above mentioned scenario I. The HRT determines the volume and capital cost
for an AD system and might be helpful in avoiding scenario II by increasing the HRT
(avoiding of washout microbial biomass and TAN) or scenario III by lowering HRT
(facilitate washout of inhibitory compounds TAN). SRT affects the volatile solids (VS)
reduction degree and thus the methane yield from biomass.
 9.4 Biogas production via anaerobic digestion | 235

Furthermore, for stable maintenance of the microbial community within the reac-
tor and thus a stable fermentation process per se other macro- and microelements (P,
S, Co, Cr, Fe, Mn, Mo, Ni, Se, and W) can be crucial, especially if industrial byproducts
(with unbalanced elemental composition) are used.
Particle size of the substrate has a direct influence on the biogas yield, since re-
duction of size improves the surface/volume ratio and correspondingly increases the
rate limiting step of hydrolysis. Increased reaction speed favors more complete sub-
strate degradation rate by given SRT and has a positive impact on the methane yield.
Nevertheless, balance between energy input (substrate size reduction) and output (ad-
ditional biogas) is often negative or neutral and does not pay off.
Despite the difficulties mentioned above, huge experimental and applied knowl-
edge of the process has been gained in recent years. Efficiency by the conversion of
chemically bound energy (biomass) into gaseous fuel (methane) is high. Theoretically,
most of the energy (up to 90%) can be converted into methane, whereby remaining en-
ergy is used for maintenance, metabolic activity, and de novo synthesis of microbial
community. This high efficiency of energy conversion and comparably low expendi-
ture on equipment complexity by the application led to a proliferating industry in re-
cent decades. The degree of application is growing worldwide, not least because of
the need for alternative energy sources as replacement for fossil fuels.
Biogas (Table 9.1, Table 9.2) presents a suitable energy source, which can be used
for electricity generation via CHP (combined heat and power), car fuel (established
technology), or launched into the gas grid. For direct use in CHP motors the biogas has
to be dry and free (< 0.15% v/v) of H2 S gas in order to minimize corrosion effects. Purifi-
cation of biogas from H2 S is usually carried out by a natural biological desulfurization
process, which takes place when oxygen is supplied. For the supply into the natural
gas network the methane content has to be increased to min. 89% v/v (whereby max.
6% CO2 and 5 H2 are tolerable). For the separation of contaminate gases (mainly CO2 )
different methods are successfully applied (e.g., pressure swing adsorption, chemi-
cal absorption [CO2 scrubber], pressure water wash, and membrane separation pro-
cesses).
The versatility of applications of biomethane led to creation of specialized facil-
ities (biogas plants), which not only digest residual biomass but growt extra plant
material (energy crops) for biogas generation (Table 9.3).

Tab. 9.1: Characteristic biogas composition and theoretical CH4 yields of different substrates.

Substrate CH4 % in biogas CO2 % in biogas Theoretical/maximal methane yield LN kg−1

organic dry weight
Carbohydrate 50 50 415
Proteins 60 40 446
Fats 72 28 1014
236 | 9 Continuously operating bioprocesses – production under steady state conditions

Tab. 9.2: Characteristic biogas and CH4 yields of commonly used energy crops.

Substrate Biogas yield LN kg−1 CH4 % in biogas Methane yield LN kg−1 organic
organic dry weight dry weight
Maize silage 450–700 52 234–364
Maize cob 620–850 54 335–459
Sugar beet 800–860 53 424–456
Grass silage 560–620 54 302–335
Sunflower 420–540 55 231–297
Wheat grain 700–750 53 371–398
Rye grain 560–780 53 297–413
Red clover 530–620 56 297–347
Fodder beet 750–800 53 398–424

Tab. 9.3: Process parameter of a full scale biogas plant fed with energy crop silage and manure at
two different organic loadings rates.

Organic HRT/SRT Biogas pro- Biogas yield Methane yield VS degradation

loading rate [d] ductivity [m3N kg−1 VS] [Nm3 CH4/kgVS] rate [%]
[kg m−3 d−1 ] [m3N m−3 d−1 ]

2.11 130 1.5 0.73 0.40 88

4.25 75 2.91 0.69 0.36 83

Typical process parameters of a full scale biogas plant are shown in Table 9.3, whereby
two different organic loading rates are applied. Higher OLR leads to much higher vol-
umetric biogas productivity (almost two fold). However, higher OLR has a direct de-
creasing impact on HRT/SRT, which leads to a slight decrease in biogas and methane
yields on a VS basis, which for its part is also reflected in a lower degradation rate.
In some countries the application degree is already reaching the natural limi-
tation regarding substrate reinforcements. Here, intensive research is ongoing in or-
der to find alternative regenerative substrates for anaerobic biogas generation. A very
promising alternative might be the use of microalgae biomass, which has higher areal
productivity than plant material (see Chapter 8 on microalgae). Nevertheless, high in-
vestment costs in algae cultivation plants and their natural ability to resist (partially)
anaerobic fermentation processes prevent this technology from large scale applica-
tions nowadays.

9.5 From minerals and particles – technical concepts for using

chemolithotrophic microorganisms

In Chapter 3 we learned that culture media consist of components for biomass and
organic product formation. For the most part, this includes a set of standard elements
 9.5 Technical concepts for using chemolithotrophic microorganisms | 237

Tab. 9.4: Most important examples of inorganic redox reactions in microorganisms and technical
applications.

Reaction Reaction pattern Group / genera Technical Remark

microorganism process

Nitrification 2NH3 + 3O2 → 2NO−2 + 2H+ + 2H2 O Nitrosomonas Wastewater Anaerobic

2NO−2 + 2O2 + 2H+ → 2NO−3 + 2H+ Nitrobacter
Denitrification 2NO−3 + 10e− +12H+ → N2 + 6H2 O Pseudomonas Wastewater Anaerobic
Intermediates NO−2 , NO, N2 O Paracoccus
denitrificans
Sulfur H2 S → HS− → S0 → S2 O−2
3 Thiobacillus Bioleaching Mainly
oxidation aerobic
Sulfate SO−2
4 → S → H2 S
0
Desulfuromonas Side- Anaerobic
reduction Organics or H2 as electron donor Desulfobulbus reaction
Desulfobacter in anaerobic
digestion
Iron Fe−3 → Fe−2 + e−1 Thiobacillus Also other
oxidation Gallionella metals
Leptothrix
Hydrogen H2 + O2 → H2 O + e− Hydrogenobacter Power to fuel Microaero-
oxidation H2 + CO2 + CO → H2 O + e− philic

(C, O, H, N, P, S), which are balanced with the stoichiometry of the organism. Indepen-
dently of the species to culture, these elements always make up the substantial part
in biomass. For redox reactions gaining energy to drive the metabolism, substrates
have to be consumed in excess and products others than biomass are formed. Both
of them contain mainly C, O, and H. Aerobic heterotrophs use organic compounds as
substrates to be oxidized and oxygen as end electron acceptors on the one hand and
CO2 and H2 O as products on the other.
One particular group of microorganisms uses inorganic substrates in redox reac-
tions for energy generation: the so called chemolithotrophs (see Chapter 2). These or-
ganisms are taxonomically very diverse and developed manifold strategies to survive
in habitats with low organic nutrient concentrations. Using the energy gained from
the redox reactions many groups are able to use CO2 as carbon source and are then
called chemolithoautotrophs. This ability allows chemolithotrophs to even populate
terrestrial subsurfaces 3 km below sea level. Chemolithotrophs play a crucial role in
nature since they participate in the biogeochemical cycling of nitrogen and sulfur, and
the formation of soil from inorganic material. Consequently, technical applications are
already in use or under investigation (Table 9.4).
The capability of chemolithotrophs to oxidize inorganic material has been known
for a long time. It is therefore not surprising that they are industrially utilized. Relevant
bioprocesses according to the redox reactions in this list are shortly outlined in the
next paragraphs.
238 | 9 Continuously operating bioprocesses – production under steady state conditions

Ammonia oxidation is an important step in the natural N cycle, enabling plants

to take up nitrate as N source. Furthermore, it is a crucial step in wastewater treat-
ment. Ammonia e.g., from municipal sewage is converted to nitrate. This process step
is called nitrification. Wastewater treatment based on chemolithotrophs is therefore
one of the most important bioprocesses in terms of material conversion. The next step
in wastewater plants, the so called denitrification, is the microbially facilitated reduc-
tion of nitrate to molecular nitrogen, which evaporates into the atmosphere. Other-
wise, nitrogen would contribute to eutrophication of the environment. The nitrogen
in wastewater mainly originates from artificial fertilizers. Plants incorporate the ni-
trogen and are processed into human food. The nitrogen ultimately finds its way into
wastewaters in the form of urea and protein. Considering the huge amount of energy
for ammonia production from atmospheric nitrogen by the Haber–Bosch process (1%
of world energy consumption) denitrification is a huge waste of resources. A much
more reasonable way of nitrogen disposal could be to use wastewater derived nitrate
directly as fertilizer. An obligatory prerequisite for this purpose is the efficient sepa-
ration of nitrogen from other wastewater pollutants. One option to do so is the culti-
vation of microalgae. Microalgae are one of few organisms able to incorporate diluted
nitrogen and phosphorous compounds in their biomass.
Syngas (short for ‘synthesis gas’) is a gas mixture consisting of H2 , CO, and
CO2 . It is produced e.g., from renewable hydrocarbon feedstocks by steam reform-
ing (Fischer–Tropsch process). In chemical engineering it is used for the synthesis of
large organic molecules or as fuel for combustion engines. Looking at Table 9.4 we
learn that microorganisms are able to perform synthesis steps on syngas as well. This
gains more and more relevance, as H2 will be available in the future from electrolysis,
where electricity is generated by sun or wind energy. However, fluid fuels will still be
necessary e.g., for aviation. Due to this strong demand current research focuses on
producing fuels or bulk chemicals by syngas fermentation. The metabolic pathways
in biological conversion show a relatively high product specificity e.g., for alcohols
like ethanol or butanol or organic acids like acetate. In this regard, it is remarkable
how an efficient chemical process and a biological continuous process based on a
gaseous substrate are coupled. Nevertheless, the low solubility of H2 reduces volu-
metric productivity, which can be counteracted by higher pressure. Reported values
are up to PV = 50 mg ⋅ L−1 ⋅ h−1 .
Chemolithotrophic oxidation contributes significantly to geochemical iron and
sulfur cycling and the responsible organisms inhabit many different habitats. How-
ever, the ability to oxidize sulfide minerals was already industrially tapped in the
early industrial ages. This is important because, unless oxidized, heavy metal sul-
fide minerals are insoluble in water or acid solutions. There is historical evidence
that bacterial leaching was used for copper recovery from Rio Tinto in the eigh-
teenth century. Bioleaching today is a standard technique in hydrometallurgy, where
chemolithotrophic bacteria are used for the extraction of copper, zinc, cobalt, nickel,
and uranium from their ores. One of the major advantages over traditional mining is
 9.5 Technical concepts for using chemolithotrophic microorganisms | 239

that bioleaching works with relatively low concentrated ores with possible extraction
yields over 90%. Bioleaching is generally simpler and cheaper to operate. The oper-
ation costs for cathode copper for example range between US$0.18 and US$0.22 per
pound and can compete with traditional smelting.
There are three different bioleaching techniques, depending on the resources to
be processed. These are dump leaching for low grade ores and waste rock (Figure 9.12),
agitated leaching for high grade chalcopyrite concentrates, and heap leaching. Heap
leaching is mainly used for newly mined run-of-the-mine ores, which contain inter-
mediate grade oxides and secondary sulfides. The material can be leached during the
mining or crushed and acidified before deposition on the heap. Traditional mining
can also only be partially replaced by bioleaching, for example to replace energy de-
manding and therefore cost intensive crushing and grinding steps. This can be seen
as a continuous process, where the solid substrate and the attached cells are kept in
the reaction volume, which is indeed a very large volume even if not exactly enclosed.
The main principle of leaching is to transform the insoluble sulfides into soluble
salts, which can be collected. This involves acidophilic metal sulfide oxidizing bac-
teria and archaea that actually oxidize Fe2+ and/or sulfur compounds. This reaction
provides Fe3+ and protons, which subsequently attack the metal sulfides:
2Fe2+ + 0.5O2 + 2H+ → 2Fe3+ + H2 O (9.15)
The bioleaching process of copper is schematically depicted in Figure 9.12. The most
important copper minerals are sulfides, for example chalcopyrite (CuFeS2 ), bornite
(Cu5 FeS4 ), chalcocite (Cu2 S), and covellite (CuS). The process involves two bacteria,
Thiobacillus ferrooxidans, which catalyzes the oxidation of iron (Fe2+ → Fe3+ ), and
Thiobacillus thiooxidans, which catalyzes the oxidation of sulfur.
These sulfur oxidizing bacteria oxidize sulfides to sulfates and protons, which
keeps the pH low and therefore supports the solution of Fe ions and the solution of
metal sulfides:
Cu2 S + O2 → CuS + Cu2+
CuS + O2 → Cu2 + + SO2−
4 (9.16)
3+ 2+ 2+ +
CuS + 8Fe + 4H2 O → Cu + 8Fe + SO2−
4 + 8H
The dissolved copper ions (Cu2+ ) are then extracted, for example by ligand exchange
solvent extraction.
In the beginning of industrial bioleaching, only microorganisms were used,
which naturally occurred on the leaching site. With rising success and acceptance
of bioleaching, attempts were made to optimize the performance of the microbial
population involved in the process, for example towards acid and temperature tol-
erance. Some of the hyperthermophilic organisms used today were extracted from
natural sources like hot springs and volcanic lava, and can catalyze reactions at 65 °C
and higher. Although a large proportion of approximately 30 naturally occurring mi-
crobial strains that can be found on a bioleaching site are chemolithotrophic bacteria,
240 | 9 Continuously operating bioprocesses – production under steady state conditions

recirculaon

Thiobacillus ferrooxydans collecon ditch

copper sulfide copper sulfate

aeraon

impermeable membrane

purified soluon

copper extracon

Fig. 9.12: Dump and peripheral devices for bioleaching of copper. Microorganisms oxidize weak
soluble metal sulfides for mining. Fe3+ is regenerated by Thiobacillus thiooxidans in the collection
ditch and recirculated onto the heap.

most extremophiles belong to the archaea. Leaching sites are often inoculated with
industrial strains of microorganisms in order to ensure a high rate of microbial activity
prior to the start of the bioleaching operation.
Today, scientists are thinking one step further and numerous applications for mi-
crobial utilization of inorganic material are currently under investigation. These in-
clude extraction of rare earths from industrial slags or waste electronics, heavy metal
resorption from wastewaters, or even biological rust removal.
Besides redox reactions for the delivery of redox equivalents and energy produc-
tion some microorganisms perform inorganic reactions for other purposes. Among
these are detoxification processes, or building up skeleton elements or functional
structures for sensorial tasks. In this regard we should take a look at well known ex-
amples of inorganic particle formation. Nanoparticles of silver and gold are reported,
while less noble metals like Ni, Fe, Cu, Zn, and others are precipitated as insoluble
salts including sulfides or carbonates. Nanoparticles have very interesting qualities,
which attracted the attention of material sciences in these processes. A typical ‘nano’
attribute is that the electron mean free path exceeds the dimensions of the particle.
This leads to specific qualities different from the bulk material. An interesting example
is magnetite nanocrystals formed by Aquaspirillum (Figure 9.13).
While magnetite Fe3 O4 as such is nonmagnetic, these particles are superparamag-
netic, allowing the cells to detect the direction of the terrestrial magnetic field. Mag-
netospirillum cells inhabit the microaerobic zone in the soil. In the case of anoxia they
use this orientation signal to move in an upward direction in order to find better con-
ditions. What happens if a soil sample from the northern hemisphere is buried in New
Zealand? Please don’t try this out as it is forbidden. Magnetite nanoparticles (nowa-
 9.5 Technical concepts for using chemolithotrophic microorganisms | 241

Fig. 9.13: Electron microscopic picture of

Aquaspirillum magnetotacticum. The cell uses
magnetite nanoparticles (artificially colored in
yellow) for orientation along the Earth’s mag-
netic field (© Science Photo).

days chemically produced) are of some technical interest, as they are applied in mag-
net separation in biotechnology.
Another example of nanoparticles used in biotechnology are CdS crystals for their
strong fluorescence abilities. Cd is a highly toxic element for all organisms. Yeasts liv-
ing in contaminated soils can take up Cd and precipitate it as CdS in a detoxification
reaction. The particles are coated by a protein layer to prevent Cd diffusion inside the
cell (Figure 9.14). Such yeasts can survive in heavily contaminated environments and
can accumulate up to 20 mg Cd/g CDM, which exceeds the lethal dose for other cells
by orders of magnitude.
The mechanism works in a way that phytochelatine molecules can bind to a stoi-
chiometrically fixed number of Cd+ ions (one or two). Then they arrange themselves in
a defined three-dimensional order to cover the CdS. The technically interesting results
are very (probably to one atom accurate) monodisperse particles, which cannot be
produced in any other manner. This is also an example of spatial self-organization as
a basic and important principle in nature.

Pepde

CdS85 core
Fig. 9.14: Structure of CdS particles in yeast as the result of self-
20 A organization on the molecular level.
242 | 9 Continuously operating bioprocesses – production under steady state conditions

Fig. 9.15: (a) ESEM picture of a single coccolith of the coccolithophorid alga Emiliania huxleyi.
(b) 10–15 single coccoliths build up a coccosphere, which surrounds the cell. The function of this
particle envelope is still unknown (© KIT).

Besides inorganic particle formation for detoxification, there are popular examples
for microorganisms, which build up inorganic matter to form exoskeletons that cover
the cell surface. One of the most impressive organisms is the coccolithophorid alga
Emiliania huxleyi. This alga produces large amounts of calcite platelets, so called coc-
coliths, which are 3–5 µm in diameter and composed of several crystalline subunits
(Figure 9.15). This allows for an exceptionally sophisticated three-dimensional struc-
ture, while shape and size parameters are very narrowly distributed – a property that
is exclusively delivered by nature. Until now, there is no possibility to replicate coc-
coliths or similar particles synthetically. Another interesting difference to industrial
calcite particles is the presence of organic components. These components form an
organic matrix, which is embedded in the calcite structure and which is known to
control the crystal formation during the biomineralization process.
Coccoliths represent an innovative particle species potentially suitable for numer-
ous industrial applications. Current suggestions include all sectors of conventional
calcite applications such as bulk products like carrier material for paints and lacquers,
fillers for tablets, adhesives, and even cements. High tech applications such as semi-
conductors, lasers, optics, liquid displays, ultrafine surface modification, high quality
photo papers, and self-cleaning surfaces are also conceivable possibilities. The pro-
duction of coccoliths demands an efficient cultivation process with high calcite pro-
ductivities. This means that the alga needs large amounts of initial substrates in order
to form equivalent amounts of calcite. Unlike other media recipes, cultivation media
for coccolith production needs to contain large amounts of calcium salts like calcium
chloride or even calcium carbonate. These substrates demonstrate a completely new
challenge in media design, since calcium chloride inhibits growth in large concentra-
tions and calcium carbonate is hardly soluble in water. Besides coccoliths, there are
other phytoplankton particles, for example the silica shells of diatoms or the enor-
 9.6 Complex molecular structures – how they influence the process | 243

mously diverse skeletons of radiolaria, and their potential in industrial application is

yet to be discovered.

9.6 Complex molecular structures – how they influence

the process

Most products produced by biotechnological means are basically compounds with

natural occurrence in microorganisms, plants, and animals and are therefore prefer-
ably employed in food or as drugs. The other way round, production is not only pos-
sible by biotechnological production with microorganisms, but also by chemical syn-
thesis or by extraction from higher organisms especially plants. The decision for a
biotechnological process depends on several aspects, of which the structure of the
molecule is one. From the structure of penicillin it can easily be understood that a
complex molecule can more easily produced by a microorganism than chemically.
This holds also for the last step in production, the enzymatic cleavage of the residual.
Here at a very specific site of the molecule a substitution is introduced. Only enzymes
or cells can be so specific: this characteristic feature is called ‘regioselectivity’. Where
this is the case very strictly also the term ‘regiospecificity’ is used. As an example to
follow this aspect a bit further we will have a look at the production of vitamin C.
It is since the beginning of the twentieth century that scientists understood the
role of specific substances in food as protective agents against different kinds of defi-
ciency diseases (Casimir Funk, 1912 and later). The seaborne disease scurvy e.g., could
be effectively combated by vitamin C. These ‘vitamins’ were later identified, described,
and synthesized. The impact of these findings is so high that 12 Nobel prizes for 20 lau-
reates were awarded. In the beginning these trace nutrients were extracted from plant
material. In the case of vitamin C the human need is about 75 mg per day, usually
completely covered by fruit or vegetables. Note that paprika contains 100 times more
vitamin C than citrus fruits, it which it was first detected. Production on a technical
scale was introduced in 1934 by Tadeus Reichstein, already 20 years after its discov-
ery, a milestone in the history of nutrition. This included even a biotechnological step.
Figure 9.16 gives an overview over the so called ‘Reichstein–Guessner’ process.
1. In the first step, D-sorbitol is formed by the hydrogenation of glucose over a nickel
alloy.
2. By microbial oxidation with Gluconobacter oxydans it is possible to oxidize the OH
group at C 5 of D-sorbitol specifically and thus to generate L-sorbose. The dehy-
drogenation reaction with Gluconobacter oxydans is regiospecific and follows the
empirical Bertrand–Hudson rule. Due to this rule, polyols with cis arrangement of
two secondary hydroxyl groups in D-configuration to the adjacent primary alcohol
group (D-erythro configuration) are oxidized to the corresponding ketoses.
 * O *
OH OH

OH OH Gluconobacter O * O OH
Ni/ H2 oxydans
HO HO HO HO OH
OH OH OH
HO OH
OH 1 OH 2 HO 3
*
OH OH OH

D-glucose D-sorbitol L-sorbose α-L-sorbofuranose

Ketogulonicigenium vulgare

+Helper strain 4 Acetone H+

(Gluconobacter oxydans)

O OH O OH

OH HO OH
HO O O
+ + * O
O O H H * O
HO HO HO Pt or Pd/ O2 O
O O
* OH OH O
O O O
HO OH 7 HO HO 6 O 5
* *
OH OH

2-keto-L-gulonic acid Diacetone-2-keto-

L-ascorbic acid Diacetone-L-sorbofuranose
Endiol-form keto-form L-gulonic acid
244 | 9 Continuously operating bioprocesses – production under steady state conditions

Fig. 9.16: Production steps of the Reichstein–Güssner process employing protective groups and modern shortcuts by microorganisms.
 9.6 Complex molecular structures – how they influence the process | 245

3. The open-chained form of L-sorbose is in equilibrium with the cyclic hemiketals

L-sorbofuranose (α and β form) and L-sorbopyranose (α and β form).
4. In an acid-catalyzed reaction of α-L-sorbofuranose with acetone 2,3,4,6-di-iso-
propylidene-α-L-sorbofuranose (‘diacetone-L-sorbofuranose’) is formed. This in-
sertion of the protective groups is necessary to enable the selective oxidation of
the terminal OH group of L-sorbose. Otherwise, all hydroxyl groups are available
for oxidation.
5. Diacetone-L-sorbofuranose is then oxidized over a Pt or Pd catalyst to di-O-iso-
propylidene-2-ketogulonic acid.
6. The acetal protecting groups are removed by treatment with acid.
7. The keto form of the resulting 2-ketogulonic is due to keto-enol tautomerism in
equilibrium with the endiol form. It rearranges in acidic conditions to L-ascorbic
acid. In an intramolecular esterification the OH group in position 4 reacts with the
carboxyl group spontaneously under elimination of water to a γ-lacton.

The important step is the regioselective transformation from sorbitol to sorbose. This
partial oxidation is interesting insofar as sorbitol is a nearly symmetric molecule. A
chemical catalyst could hardly distinguish between the two sides and would deliver
many different isomers. So a chemical synthesis would be much more complicated
as can be seen in the following steps where protective groups have to be introduced.
Here high regioselectivity of biosystems shows a clear advantage over chemical syn-
thesis. This process has been standard for many years and is in operation even today,
although the final yield is only 50% and accompanied by the need for solvent recy-
cling, high temperatures, and high pressures in some of the steps.
The story shows how fast biotechnological realization can be, after scientific
recognition is achieved. Politics also play a role. In 1994 the World Bank published a
paper on the importance of vitamin supplementation of food for overcoming vitamin
malnutrition. The Chinese government declared vitamin production a key technology.
This encouraged further research and led to today’s dominance of Chinese research
and production of vitamin C. The target of new processes is of course, as becomes
clear from the sketch in Figure 9.16, the substitution of the protective group steps by
a direct partial oxidation sorbose directly to 2-KLGA with another microorganism,
Ketoguloronicigenium vulgare. This only works together with a ‘helper strain’, e.g.,
Bacillus megaterium for not completely understood reasons. This modern two step
process delivers up to 130 g ⋅ L−1 vitamin C with a yield of 80% related to sorbitol.
Further improvement by different process variations and other genetically engineered
microorganisms is in progress. Today (2014) the world production amounts to 110,000
tons per year mainly in China and is sold in a competitive market for approximately
US$6–8/kg. Western producers serve a premium market segment protected by quality
brands. Around 50% goes into the pharmaceutical market, 25% is sold as antioxidant,
while 15% is uses as additive in beverages.
246 | 9 Continuously operating bioprocesses – production under steady state conditions

All these achievements would not be possible without the ability of Gluconobacter
and Ketoguloronicigenium for performing partial oxidation. So a short look is advis-
able. The enzymes involved are mainly located in the periplasm, so the substrate does
not have to enter the cytoplasm and the product leaves back to the environment. Most
of these dehydrogenases are characterized by a remarkable regiospecificity but rela-
tively low affinity, and also by a broad substrate range. This makes a two step process
necessary to avoid direct contact of K. vulgare with sorbitol possibly leading to un-
wanted byproducts. Both stages are batch processes with product titers above 200 g/L
for sorbose and 100 g/L for 2-KGA. As oxygen transfer is comparatively low, airlift re-
actors are chosen. Substrate overshoot metabolism is not observed. In fact, the reac-
tion pattern reminds one of an overshoot metabolism itself, as the TCC is incomplete
and could be bridged by the NADPH gained from the uncomplete oxidation. To ease
growth of the cells a rich medium with high titers of corn steep liquor and yeast extract
is used. Nevertheless, careful control of metabolic turnover rates could finally lead to
a one step process provided G. oxidans works faster, keeping the sorbitol titer low.
Besides further improvement of the bacterial route via genetic engineering, we
can also think of alternatives, e.g., enzyme reactor with NADPH recycle, or complete
synthesis by plants cell culture or microalgae. This is actually possible, but product
titer of 2 g/L makes downstream processing including cell disruption for the time being

Tab. 9.5: Unique selling points, molecular structure, and process.

Aspect Substance class Examples products Process

Enantioselectivity Glutamate Direct synthesis by
Stereospecificity Corynebacterium
Enzymatic
Regioselectivity Penicillin Direct synthesis
Enzymatic substitution
Vitamin C Partial oxidation
Extracellular
Macromolecules Proteins Technical proteins Direct synthesis,
Sequential order extracellular, Bacillus
Hormones Intracellular e.g., E. coli
Antibodies Animal cell culture
Polysaccharides Xanthan Direct bacterial synthesis
Schizophyllan extracellular
Nucleic acids RNA therapeutics Direct synthesis
DNA computer (?) intracellular
Fatty acids Health food Direct synthesis
Lubricants
Autocatalysis Whole cells Baker’s yeast Fed batch
Starter cultures Batch
 9.7 Current discussion about continuous processes – motivation and obstacles | 247

too expensive. Even genetically engineered higher plants cultivated in greenhouses

are not excluded. Other examples for incomplete oxidation are vitamin B2 (precur-
sor D-ribose with Bacillus), vitamin B12, biotin (possible), or regiospecific oxidations
of progesterone by Rhizopus nigricans leading to hydroxy-progesteron and finally hy-
drocortisone. Further and more general examples with some insight into the relation
between molecular structure and process are collected in Table 9.5.

9.7 Current discussion about continuous processes – motivation

and obstacles

Continuous processes are for the time being applied for large amounts but small
molecules, nevertheless not in the extent as could be expected or envisaged when
summing up the advantages and obstacles (Table 9.6).
Especially in the pharmacuetical industry there are some reservations but with a
positive trend for more applications. Here are statements from a representative from
industry:

Today’s reality follows the ‘legend’ that continuous bioprocesses are more for academics (‘conti
cult’) pretty closely, although there are quite some approved processes that use continuous (per-
fusion) cultivation processes for an extended period of time. For those processes, genetic stability
has been demonstrated during the entire perfusion process. For the downstream process, things
seem to be moving now. In today’s approved processes, all DSP operations are indeed batch-
wise. New technologies have been developed that would allow transforming the DSP steps into
an integrated continuous biomanufacturing process. This includes single-pass tangential flow
filtration and multicolumn chromatography. Although some of these technologies may require
slightly more complex equipment, the process itself should be less complicated because (a) they
are operating at a steady state, (b) they provide better segregation between different process flu-
ids, and (c) the impact of dead volumes is far less significant in continuous processes than in

Tab. 9.6: Pros and cons of continuous processes.

Physiological aspects Technical aspects

Cells are fully adapted +++ Reactor always full at high biomass +++
concentration
Insidious contamination −− No batch-wise quality control possible −−
Creeping loss of plasmids −− ++
High product concentration and pro- − Can be compensated by cell recycle +
ductivity not simultaneous
No high dilution rates for product − Can be compensated by reactors in series +
inhibition
Substrate lost in effluent − Low reactor downtimes +
Continuous downstream desirable +/−
248 | 9 Continuously operating bioprocesses – production under steady state conditions

batch process. The potential advantages of a downstream processing platform are improved pro-
ductivity (producing more product in smaller process equipment, thereby enabling a fully dispos-
able process), enhanced process control, and with that better product quality control. This last
point has also inspired regulators (e.g., FDA) to support the trends towards continuous process-
ing in pharma and biopharma. So – in short – yes: the current opinion still follows the mentioned
legend, but things are rapidly changing. I would hope that academic institutions, who educate
tomorrow’s process engineers and process scientists, would appreciate the current state of the
art (batch) but also address the trends that are natural in a maturing industry (which includes
implementation of continuous processing).

In scientific social networks the topic is further discussed; links are given in the refer-
ence list.

9.8 Questions and suggestions

1. Is it possible to increase the maximum biomass productivity in the standard

example by decreasing substrate affinity e.g., by genetic engineering?
2. Calculate the specific substrate uptake rates in the data example of contin-
uous ethanol production in Figure 9.5 and discuss the influence of ethanol
concentration on kinetics.
3. You produce a recombinant protein with a genetically engineered cell line
of D = μ eng = 0.2 h−1 . By an occasional reverse mutation 1% of the cells
are wild strain cells in the reactor. They can growth faster without metabolic
burden (higher yield) μwild = 0.3 h−1 . Calculate the time until they are 10%
and finally overgrow the mutant.

1. D ⋅ kS
PV,X (D) = D ⋅ c X (D) = y X,S ⋅(cS,f − cS (D)) = y X,S ⋅(cS,f − ) (9.17)
y X,S ⋅ rS,max − D
For the optimum the first derivative with respect to D is zero:

PV,X (D) kS
= 0 → Dopt = rS,max ⋅ y X,S ⋅ (1 − √ ) (9.18)
dD k S + cS,f

As long as cS,f ≫ k S no significant improvement is possible. For lower D values

a significant increase in process yield is not possible. Find the discussion of the
solution in the simulation file of the supplementary material.
2. The specific substrate rate can be calculated from the substrate balance and the
measurements:
D ⋅ (cS,f − cS )
D ⋅ (cS,f − cS ) − rS (cS ) ⋅ c X = 0 → rS ≈ (9.19)
cX
The results are plotted in Figure 9.17.
 9.8 Questions and suggestions | 249

11 1.0

10
0.8
9
rS [g⋅g-1⋅h-1]

D [h-1]
8
0.6

0.4
6

5
0.2
4

3 0.0
10 15 20 25 30 35 40 45 50
Ethanol cp [g/L]

Fig. 9.17: Specific growth rates as a function of substrate and ethanol concentration.

At high ethanol concentrations rS and μ are low but increase with increasing sub-
strate concentrations. When substrate is in saturation further increase relates to
declining ethanol concentration. So cEth has an impact on rS and on y X,S .
3. Setting up the dynamic balance equation for the wild type:

dc X,wild
= μ wild ⋅ c X,wild − μ eng ⋅ c X,wild
dt
The wild strain grows up in the reactor with an apparent growth rate of μapp =
μ wild − μ eng .
10 Measuring principles – how to put an end
to the blind flight
Nobody would enter an airplane knowing that the pilot can control the flight only
by a view out of the cockpit window. Bioprocesses are very complex as well requir-
ing detailed and continuous monitoring and technical specifications for the course of
the process. This chapter will give an overview of the technical possibilities, potential
benefits, and constraints with respect to employment of at least basic measurement
devices. The direct physical principles generating the relevant signals will be outlined
and followed by the measurement chain up to a level where the meaning for the pro-
cess can be evaluated. This gives the basis for a generalized view of the structure of
sensors for applications in complex environments.

10.1 Only a look through the keyhole – what measurement means

for process understanding and optimization

One of the major advantages of bioreactors over shaking flasks or roller bottles is the
possibility to measure and control many environmental parameters influencing the
microbial growth process. Even slight deviations from optimal values can change the
specific growth rate or specific production rate by several percent. Microorganisms
change many of these parameters by their metabolic activity. The other way around,
changes of pH or pO2 for example offer useful information to assess the current physi-
ological state of the cells. Targeted changes in environmental parameters are a means
for optimized process strategy. For the production of (bio) pharmaceutical products a
transparent and comprehensive documentation of each process run is demanded by
the authorities. Measurement data are required to prove that all of the important pa-
rameters are within a prescribed range to ensure product safety. The key phrase is pro-
cess analytical technology (PAT), a set of rules for how these aims are to be achieved.
Especially in technical environments there are some reservations against installation
of sensors as they are suspected to increase contamination risks. Detractors speak of
making the reactor into a ‘Swiss cheese’. Furthermore, measured values are not rigor-
ously evaluated to give hints for process enhancement.

10.2 State of the art – overview of measurement

at a standard reactor

To observe and control the environment many measurement variables are important
and usually measured. Others should be measured but no reliable standard sensor ex-

https://doi.org/10.1515/9783110315394-010
 10.2 State of the art – overview of measurement at a standard reactor | 251

qTit off-gas analysis

cTit

pH, pO2,
OD Z1 biosensors

base/acid/
p quasi-online analycs
anfoam Z2

TM, TJ off-line sampling point

X, S, P
qS,f , cS,f qout harvest
substrate

NS, IS, PS

gas inlet

VR
Air N2 CO2 O2

Fig. 10.1: A bioreactor equipped with on-line and at-line sensors for different parameters; the param-
eters to be measured are shown as rectangles, the related devices as usual flow sheet symbols.

ists. Specific sensors for dedicated processes are in use only in rare cases. In Figure 10.1
an overview of usual measurements on a bioreactor is shown (Table 10.1).

Sensors are classified according to the measurement principle but also according to the means of ap-
plication. The definitions of some technical terms are given in the following paragraphs as background
information for this chapter.

‘In-line’ measurements are based on sensors integrated into the reactor and usually
have direct contact with the medium. They need specific flanges at the reactor and
have to survive high temperature during sterilization. ‘At-line’ measurements are
mounted outside the actual reaction space but in the gas phase or in a bypass (side
stream, slip stream). Here size or technical design are not so limited, but measurement
is more indirect. The terms in-line and at-line relate therefore to the spatial structure of
the process. In cases where a sensor measures a compound but without direct contact
with the medium or the gas phase the term ‘noninvasive’ is used.
Another classification of terms is connected with measurements related to tempo-
ral resolution. Most in-line sensors deliver values in rapid succession. That holds also
for some at-line sensors. So the last value represents the current state of the measured
variable. Such sensors are called ‘on-line’ sensors. According to the Nyquist–Shannon
sampling theorem the measuring frequency should be at least ten times faster than
the characteristic time constant of the process. The maximum frequency is limited by
the establishment of the measured variable and the physical state of the sensor. The
252 | 10 Measuring principles – how to put an end to the blind flight

Tab. 10.1: Labels, names, and some additional information on standard sensors for fermentation.

Label Name Physical principle Meaning for process

TM , TJ Temperature Change of electrical resis- Dedicated optimum
(medium, jacket) tance, PT100
P Pressure Mechanical force by pressure Drives exhaust gas flow, con-
difference trol during sterilization
pH pH value Electrochemical cell Dedicated optimum
pO2 Oxygen partial Electrochemical cell Minimum to be respected
pressure
OD Optical density Light scattering Estimation of biomass con-
centration
N S , IS , PS Speed, electrical Electromagnetic Power input for mixing and
current, electrical gas dispersion
power of stirrer
VR Working volume Weight via balance
Fliquid Mass flow, liquid Differentiation of balance Medium supply, balancing
signal, conversion to volume
by calibration
Fgas Mass flow, gas Hot-wire anemometer, conver- Balancing gas phase
sion to volume by calibration
xO2 , xCO2 Molar fraction gas IR for CO2 and paramagnetism Gas transfer, physiological
for O2 state
K, rH Specific conductiv- Electrochemical sensor Sum parameter for dissolved
ity, Redox potential ions

term off-line measurements in bioprocess engineering refers to sampling and analysis

in the lab. So the results are typically not available during the process. To overcome
this problem some analytic devices can be connected to the reactor by an automated
sampling system. These measurements are referred to as ‘quasi-online’. While it takes
several minutes to get a new value it is fast enough to make decisions while the process
is running.
There are other terms that are in use when designating the physical working prin-
ciple. ‘Optical’ sensors use the interaction between light and the analyte, while ‘elec-
trochemical’ sensors are based on an electrochemical effect as described below. Here
‘potentiometric’ (also ‘galvanometric’) means an electric potential and ‘amperomet-
ric’ (sometimes ‘polarographic’) an electric current as the primary measuring signal.
In contrast to ‘analog’ sensors, ‘digital’ sensors contain an analog/digital converter
(AD converter) to reduce signal corruption by noise in rough environments. ‘Software
sensors’ combine different hardware sensor signals and process the data with com-
putational methods to elucidate new information. They calculate a virtual measure-
ment like the specific growth rate μ from real measurements like off-gas analysis. It
 10.3 Physical parameters – adaptation from other fields of process technology | 253

makes process monitoring more convenient. An example is the integration of dc X /dt =

μ ⋅ c X = y X,O ⋅ OTR using the stoichiometry between cell growth and oxygen uptake.
This gives estimation of biomass as:

c X,est = c X,0 + y X,O ⋅ ∫ OTR ⋅ dt (10.1)

and
OTR
μest = (10.2)
c X,est
A sensor’s sensitivity indicates how much the its output changes when the input quan-
tity being measured changes

10.3 Physical parameters – adaptation from other fields

of process technology

Temperature is a central environmental parameter. In Chapter 2 it was already shown

that a deviation from the temperature optimum by 1 °C can cause a loss in growth rate
of several %. Measurement is a standard done by a PT100, a platinum wire of 100 Ohm
electrical resistance having a linear resistance/temperature kinetics.
Pressure in the headspace has to be measured firstly for keeping 2 bar during
sterilization and secondly during cultivation for driving the exhaust gas through
the off-gas tubes. In pressure sensors a pressure difference leads to deformation of a
membrane, which is secondarily transduced by e.g., piezoelectric effects to an electric
signal.
To measure the working volume of the reactor during cultivation, which is the
amount of medium, is not simply measurable by a float. The signal would be corrupted
by foam, the vortex induced by stirring, and last but not least by the changing gas
holdup (bubble volume) in the reactor. The only solution is to put the whole reactor
on a balance, which is precise enough to follow changes of weight during the process.
Similar reasons hold for employing balances for measuring the amount of acid, base,
and antifoam fed into the reactor. The related fluid mass flows are calculated from the
balance signal by differentiation and pump calibration.
Gas mass flow measurement is foreseen only at the gas inlet. It is based on hot-
wire anemometry. An electrically heated wire at constant temperature is overflowed by
the gas. Heat transfer from the wire to the gas measured by electrical heating power
depends on gas flow and molecular mass. So calibration has to be done for the gas
composition in use. Although mass flow controllers as the standard device to control
gas mass flow are quite expensive one should give precedence to them over the classic
rotameters.
Off-gas composition is basically a chemical parameter but based on physical ef-
fects. Standalone measurement devices are adapted from other industrial applica-
tions, as are most of the physical sensors. A specific feature of CO2 is the strong light
254 | 10 Measuring principles – how to put an end to the blind flight

absorption in the infrared (IR) range. This is commonly known from the global green-
house effect but employed here for measurement purposes. IR light is absorbed by the
gas sample, which is consequently heated up. The pressure difference to a comparison
sample is the final physical parameter to be calibrated to xCO2 and displayed. This is
an example of a measurement chain where the signal carrying the measurement in-
formation is several times transformed before it reaches the data acquisition system.
A unique characteristic of oxygen is its paramagnetic property. A gas sample ex-
periences a force in a strongly inhomogeneous magnetic field into the direction of in-
creasing field strength. This induces a measurable deviation of the gas flow or the
torsion of a dumbbell shaped element carrying a comparison sample at the other end.

10.4 Chemical parameters – employment of electrochemical

effects

Chemical parameters can be measured in the suspension but also in the gas phase. As
a first target we have a look at dissolved ions, which are important as medium com-
pounds and influence the pH value. To understand the physical basis of related sen-
sors an excursion to electrochemical effects generating measurement signals is useful.
Three basic arrangements are shown in Figure 10.2.
In the first arrangement (Figure 10.2 (a)) a solid metal electrode is in contact with water
or a salt solution, which could be the medium of a bioprocess. Metal ions will diffuse
into the medium. Thereby an electrical potential is built up, which exerts an electro-
static force (Coulomb force) on the ion in the opposite direction to diffusion until an

Me Me MeAn + -

Na+

Me+z K+
An-
semipermeable
membrane An-

buffer buffer
soluon soluon

(a) (b) (c)

Fig. 10.2: Basic principle of electrochemical effects; the potential at the electrodes depends on the
ion concentration, (a) metal electrode, (b) metal-oxide electrode. Semipermeable membranes (c)
build up a transmembrane voltage.
 10.4 Chemical parameters – employment of electrochemical effects | 255

equilibrium is reached:

Me+z (solute) + z ⋅ e− (solid) 󴀕󴀬 Me(solid) (10.3)

The resulting electrical potential is given as:

R⋅T
ϕ = ϕ∘ + ⋅ ln aMe+z (10.4)
z⋅F
This equation is called the Nernst equation according to Walther Nernst (Nobel Prize
Chemistry in 1920). The activity is defined as an activity constant γ – for simplicity here
set to 1 for low ion concentration – multiplied by cIon /c0 , where c0 is the unit 1 mol/kg.
Summarizing, this arrangement acts as an ion sensitive electrode and enables us to
measure cation concentrations. See a calculation example under questions.
In the second arrangement (Figure 10.2 (b)) the electrode is covered by a layer of
a metal salt e.g., metal oxide or metal chloride. At the inner interface a dissociation
process generates electrons e− :

MeAn(solid) + e− (solid) 󴀕󴀬 Me + An− (solid) (10.5)

At the outer interface again diffusion and electrostatic forces come to an equilibrium:

An− (solid) 󴀕󴀬 An− (solute) (10.6)

In this case the Nernst equation reads:

R⋅T
ϕ = ϕ∘ − ⋅ ln aAn−z (10.7)
z⋅F
By this ion sensitive arrangement obviously anion concentrations can be measured.
In the third arrangement a semipermeable membrane separates two compart-
ments filled with salt solutions, e.g., KCl in compartment I (left) and NaCl in compart-
ment II (right). Assuming that only K+ can diffuse through the membrane, some K+
ions will follow the concentration gradient until they are stopped by the self-induced
electric field. In this case the Nernst potential becomes:
R⋅T a K+I
ϕ= ⋅ ln (10.8)
z⋅F a K+II

This membrane potential plays an important role in biology. Nearly all cells exhibit a
membrane potential over the cell membrane. Biological receptors often work by mod-
ulation of the membrane potential. In the case of neurons the potential can be more
than 100 mV. Also, where no membrane is present, an initial gradient in ion concen-
tration can lead to an electric potential, because different ions diffuse with different
velocities. In cases where they have different charge, a diffusion potential is built up.
As an example for the application of electrochemical sensors, we will have a closer
look at pH, which is one of the most important environmental parameters to control
256 | 10 Measuring principles – how to put an end to the blind flight

in a bioreactor. Starting with a correct pH value at the beginning of the process the
hydrogen ion concentration (correctly speaking oxonium H3 O+ ) is subject to change
due to activity of the microorganisms. They may produce acids or take up acid or ba-
sic salts. One candidate is NH+4 where the uptake as ammonia leaves a proton in the
medium. In most cases, the pH value drops during a cultivation, making titration by
a base necessary. This could again be ammonia hydroxide. As long as the biomass
contains a constant quota of nitrogen and does not produce acids, ammonia dosage
by pH control is proportional to biomass growth. Here we have another example of a
popular software sensor, where the growth is estimated via ammonia dosage.
To measure pH, in principle a hydrogen gas electrode according to arrangement
Figure 10.2 (a) would be necessary. However, that would be very impractical. At this
point an observation, made firstly in the early nineteenth century, helps. It says that
protons can diffuse into glass material generating in this way a potential between the
glass surface and the surrounding medium. This effect is made possible by the small
size of protons and the ion exchange properties of glass, where Na+ ions, or Li+ in mod-
ern versions, can compete with H+ ions for the negatively charged SiO− compounds in
the glass. ‘Glass electrodes’ using this effect are probably the most widespread electro-
chemical sensors. A commercial pH sensor additionally provides all necessary parts
to form a galvanic element as a closed electrical circuit, where the voltage generated
by the protons can be measured. A schematic drawing of the tip of such an electrode
is shown in Figure 10.3.
The central part is the bulb shaped glass ‘membrane’ (about 0.5–1 mm thick) car-
rying inside and outside the pH sensitive layer of a hydrated ‘gel’ (about 10–100 nm
thick). Increasing the surface of the membrane is necessary to reduce the electric re-
sistance. In between the two sensitive layers Li+ ions (Na+ in older versions) take over

reference electrode buffer

diaphragm

internal working
electrode

pH-sensive glass
membrane
buffer

Fig. 10.3: Tip of an electrochemical pH electrode; the two compartments contain buffer, the working
and the reference electrode are visible as well as the glass membrane (sensing gel layer) and the
diaphragm.
 10.4 Chemical parameters – employment of electrochemical effects | 257

the charge transport. In principle, the electrical potential φ on both sides of the mem-
brane has to be measured. The difference Umeas = ∆φ is the voltage as the primary
signal. The cylindrically shaped electrode shaft consists of two separate annular cav-
ities forming two electrochemical half-cells corresponding to these two potentials. In
the inner volume of the probe an Ag+ /AgCl electrode – the ‘working electrode’ – is
in contact with the inner surface of the membrane via a KCl solution (0.1–1.0 molar).
The potential at the Ag+ wire is therefore the sum of the outer membrane potential,
the cc potential, the inner membrane potential, and the potential at the buffer/AgCl
interface.
The potential outside of the glass membrane is held on ground potential as the
reactor and the medium are earthed. However, this is not defined precisely enough
and has to be measured inside the medium as close as possible to the place of the
glass membrane. This is the task of the outer reference electrode. In fact, the medium
is in electric contact with the outer annular volume of the probe via a perforated ce-
ramic plate, the so called diaphragm, or via small pores (gel electrolyte). The reference
electrode, again Ag+ /AgCl, measures the potential in the outer annular volume deliv-
ering in this way a measure for the potential at the outer gel layer in contact with the
medium. The outer volume is filled with a KCl solution of the same concentration as
the inner volume. This salt is chosen as K+ and Cl− exhibit similar mobility thus avoid-
ing a diffusion potential in the diaphragm. Finally, an amplifier closes the circuit. As
the probe consists of two single electrodes, it is called a ‘combination electrode’. The
course of the potential along the circuit is given in Figure 10.4.

reference electrode outer inner working electrode

gel layer

medium
buffer
soluon

ΔEmeas ΔEElect
0
ΔEElect

ΔEprim
E
ΔEasym

Fig. 10.4: Course of potential E along the measurement chain in a pH electrode; E prim primary mea-
suring effect (approximately 60 mV/pH), E meas measured value (appr. 60 mV/pH), E elect electrode
potential, and E asym asymmetry potential (a few mV) containing also the diffusion potential through
diaphragm (approximately 2 mV).
258 | 10 Measuring principles – how to put an end to the blind flight

Fig. 10.5: Three-quarter section of a pH sensor.

The potential of the outer gel layer gives a logarithmic measure of the pH value accord-
ing to the Nernst equation. As pH itself is a logarithmically defined quantity, a linear
relation between voltage and pH of approximately 60 mV/pH is provided, this value
being the sensitivity of the pH electrode. All other potentials along the measurement
chain are constant but require a zero point adjustment and a calibration prior to use.
Cross-sensitivity, especially with Na+ or K+ in the medium, may happen preferably in
the alkaline range. Temperature sensitivity (see Nernst equation) is actively compen-
sated by the amplifier based on a Pt-100 temperature sensor integrated into the pH
probe as well. A CAD view of the different volumes is shown in Figure 10.5.
The importance of knowing the oxygen partial pressure pO2 during a cultivation
has already been discussed in Chapter 5. While pH measurement is an example of a
potentiometric (galvanic) sensor, measurement of pO2 is based on an amperometric
principle. Therefore a voltage has to be actively applied to the electrode inducing sev-
eral electrochemical reactions at the electrodes. The principle is shown in Figure 10.6.
At the cathode – the working electrode (Pt or Au) – O2 is reduced to OH− :

O2 + 2H2 O + 4e− → H2 O2 + 2OH− + 2e− → 4OH− (10.9)

At the anode – the reference electrode (Ag) – electrons are produced at the cost of the
charge of Cl− :

4Ag + 4Cl− → 4Ag+ + 4e− + 4Cl− → 4AgCl + 4e− (10.10)

The electrons close the charge balance via the electrodes and the external ampli-
fier. Consequently, the current Imeas is directly proportional to the number of oxygen
molecules being oxidized at the cathode. Employment of the two electrodes in direct
contact with the medium would induce many other electrochemical reactions. So we
need an additional element that is selectively permeable for oxygen and prevents ions
from diffusing into the space inside the sensor. As oxygen is quite hydrophobic such a
separating membrane can be selected out of a range of hydrophobic materials; Teflon
is a usual choice. Remember that also for bubble free aeration of animal cell culture
 10.4 Chemical parameters – employment of electrochemical effects | 259

polarizaon potenal

+ -

Au cathode Ag anode
working electrode reference electrode

buffer soluon
(K+ Cl- )

semipermeable membrane
O2
O2

Fig. 10.6: Measuring principle of a Clark electrode.

Teflon membranes are employed. The electrode is called a ‘Clark electrode’ in honor
of its inventor Leland C. Clark (development around 1956–1962).
After switching on the sensor oxygen inside the probe is used up, while fresh oxy-
gen diffuses through the membrane. Accordingly, a current is observed increasing with
increasing voltage. pO2 sensors are designed in a way that at around 0.6–0.8 V – the
typical working voltage – the oxygen concentration in the probe drops to zero and only
fresh oxygen diffusing through the membrane is reduced at the cathode making the
current proportional to the oxygen concentration in the medium:

Imeas ∞QO2 = k Diff,O2 ⋅ (cO2,medium − cO2,sensor ) ≈ k Diff,O2 ⋅ cO2,medium (10.11)

Higher voltage (> 1.6 V) would lead to hydrolysis of water inside the probe. As the dif-
fusion coefficient k Diff,O2 [m2 ⋅ s−1 ] may be subject to change the sensor has to be cal-
ibrated against a solution with known cO2 . That can be provided by saturation of a
medium sample with air, nitrogen (zero value) or other gases with known composi-
tion. While the physical measurement principle is based on concentration usually cal-
ibration is done on the basis of partial pressures of the calibration gases. This has to be
considered in cases where oxygen solubility changes e.g., by changing ion concentra-
tions or pressure changes during cultivation. The Teflon membrane has to be changed
from time to time as proteins can be precipitate on it especially during sterilization.
Furthermore, it has to be noted that the ions are not balanced and the electrolyte has
to be changed after a number of hours of operation.
Electrochemical sensors have some disadvantages as they have to be calibrated
and cleaned regularly to avoid aging and drift. The response time is typically in the
range of up to a minute. The time constant is given by the product of the diffusion co-
efficient k L for the analyte through the membrane and the storage capacity of analyte
in the sensor volume. So the membrane surface has to be as large as possible (sensor
260 | 10 Measuring principles – how to put an end to the blind flight

red LED green LED

opcal window
sensing membrane
O protecve membrane
O2 2 O2
O2
O2
O2
medium

Fig. 10.7: Principles of optical sensing principles: the optical pO2 sensor sends excitation light
(green LED) to the dye, and emitted light (red) is detected by a photodiode. The red LED is used for
calibration purposes.

diameter, spherical shape) and the volume as small as possible. Nevertheless, there
is a need for more robust sensors with shorter response times. An interesting prop-
erty of dissolved oxygen has been found in fact that it is soluble in a polymer mem-
brane containing a luminescent Ruthenium(II) dye complex. There it acts as a lumi-
nescence quencher. The effect can be measured by irrigating the membrane with blue
light and measuring the red luminescence light (Figure 10.7). Oxygen sensors work-
ing according to this principle belong to the photometrical sensors and increasingly
replace the classic electrochemical oxygen electrodes. Even with a response time still
being around 40 s (membrane volume) the advantages are robustness, manageabil-
ity, high sensitivity especially at low oxygen partial pressure, accuracy (+/− 0.1 mg/L),
and low cross-sensitivity.
In the previous paragraph we learned about a first example of an optical sensor,
also called ‘optode’. A color reaction is also the classic litmus test for pH measurement.
This lab test gave the idea of looking to dye reactions for use in online sensors. In
modern versions dye patches are employed allowing for optical online measurement;
see Figure 10.8. The dyes are applied as a coating on the tip of a glass fiber through
which a light beam can read out the color.
Such fiber optic sensors are available for pH, pO2 , and pCO2 . Constraints for the
dyes are that they are not toxic, can be sterilized, and that they are stable long term.
The measurement principle allows for fabricating ‘microsensors’, which are quite
small so that they can be used to resolve spatial resolution of the measured parame-
ters. The response time can be less than a few seconds. Pasting the pads inside a glass
reactor or shaking flask is also possible so that they can be read out from outside.
Such noninvasive techniques have their advantages for lab work including microtiter
plates and for disposable plastic reactors.
 10.4 Chemical parameters – employment of electrochemical effects | 261

light source detector

glass fiber

Fig. 10.8: Principles of optical sensing principles:

H+ In fiber optical sensors a patch containing a dye
sensing patch
H+ e.g., with a pH dependent color, is read out by light
H+ H+
emitted by a diode – the reflected light contains the
medium information of the pH value in the medium.

Not all of the analytes we are interested in do us the favor of showing a convenient
color or electrochemical reaction. Here we have to refer to lab methods from laborious
wet chemical procedures up to employment of costly measurement devices. Results
are available often only the next day. However, it has succeeded in bringing these pro-
cedures from the lab closer to the process. A sample is automatically taken e.g., by a
small bypass stream and led through an automated device to handle the sample and
give a measurement value. These quasion-line measurements can happen within a
couple of minutes: fast enough and with high repetition rate to make use of the re-
sult for process control. The idea of at-line sensing techniques is shown for flow in-
jection analysis (FIA) as an example in Figure 10.9. A droplet of the biosuspension
is injected into a stream of a carrier fluid to be transported to several other injection
points, here to add a dye for a color reaction, which can be read out by a detector. This
scheme can be more complicated according to the related lab protocol. Other options
for at-line measurements are online high performance liquid chromatography (HPLC),

sample indicator dye

sample sample+indicator

carrier-fluid
detector

Fig. 10.9: Working scheme of flow injection analysis (FIA) as an example of an at-line measurement.
The green sample is mixed with a dye injected exactly at the spot of the sample droplet, then color
reaction has some time to occur before it is detected.
262 | 10 Measuring principles – how to put an end to the blind flight

spectroscopy, microscopy, flow cytometry, or even nuclear magnetic resonance (NMR)

spectroscopy.
As setting up at-line measurements is not always easy, the use is limited to scien-
tific projects and some industrial processes with high quality demands.

10.5 Measuring biomass – the great unknown ‘X’

Throughout this book and generally in the bioengineering community we speak about
‘biomass’ in terms of cell number or cell dry weight. Going through the catalogues of
equipment suppliers we will not find an online sensor that gives a direct signal of these
quantities. The cell is something like a diamond, which looks differently in different
light and from different angles. From measurement of the physical and chemical vari-
ables (paragraphs above) it became clear that we have to define a specific property of
cells characterizing biomass and make it different from other compounds in the sus-
pension. In a second step a specific interaction with a test signal has to be found. This
could be light, concentration differences, indicator molecules, electric fields, or me-
chanical forces. Starting from a basic engineering view some aspects are defined and
listed in Table 10.2. With a physical view the cell is a particle with a spatial structure.
Chemical reaction technology may look at it as a heterogeneous catalyst. The view of
the cell as an ensemble of macromolecules was already employed during media de-
sign. Each of these aspects can lead to different ideas for what could and should be
measured.
With this background in mind, different measuring technologies from other fields
of physics, chemistry, and those already adapted for use in chemical engineering can
be assessed and luckily new measurement effects are found and utilized for the spe-
cific purpose.

Tab. 10.2: Characteristics of living cells.

Basic view on the cell Characteristic Specific interaction Measurement technique

Particle Size Image Microscopy
Size, refraction index Light scattering Optical density
Mass, density Weight force Sedimentation, cell dry
mass
Spatial structure, cell Electrical capacity Alternating electric field
membrane
Catalyst Chemical conversion Analyte formation Analytics + software
Charge transport Electric field Electrodes
Macromolecule Light absorbance Light Absorbance Spectroscopy
Aromatic residues Fluorescence Fluorimetry
 10.5 Measuring biomass – the great unknown ‘X’ | 263

side scaering

back scaering

light source
lens Fourier lens
system sample
suspension

Fig. 10.10: According to Mie theory spheres produce light scattering patterns of concentric rings
of light intensity; large particles scatter at small angles and vice versa. The Fourier lens focuses
scattered light from different spots in the medium but with the same direction onto one ring.

Optical methods are well developed in particle technology. Here the interaction of par-
ticles with light namely light extinction by absorption and scattering caused by diffrac-
tion, refraction, and reflection are measured. This makes samples of suspensions look
turbid in contrast to molecular solutions. Unlike in fluorescence the scattered light
has the same frequency as the incident light. The basic setup to measure scattering is
shown in Figure 10.10.
Scattering depends on the wavelength and the size of the particles. Rayleigh scat-
tering occurs where particles much smaller than the wavelength of the light interact
with the light beam e.g., small molecules. Raman scattering is caused mainly by in-
tramolecular vibrations and rotations. Raman spectroscopy is frequently used in bio-
logical and chemical labs. It can be applied also to complex fermentation suspension
to measure e.g., products but mainly in a scientific context. Of higher importance for
measuring fermentation suspensions is Mie scattering describing especially the inter-
action of light with particles of similar size as the applied wavelength. The intensity
of the scattered light is a function of the cell’s optical properties and dimensions. In
principle, it is possible to resolve size distribution of cells of a particular shape from
the measured angular scattering intensity pattern. In practice, not all of the informa-
tion contained in the scattered signal is used. Especially the scattering patterns are
ignored or averaged because the detection of the transmitted signal is foreseen only
at a single light detector mounted in the direction of the incident light. This reduces
the approach to a turbidity measurement. The basic application is the so called opti-
cal density, which is the standard approach for off-line determination of biomass in
a spectrometer. Assuming a light beam with a known intensity ILight,0 travels through
a suspension of biomass along an optical path with length dPath an attenuated light
intensity ILight (c X ) can be measured. Usually, the Lambert–Beer law is employed for
264 | 10 Measuring principles – how to put an end to the blind flight

Tab. 10.3: Measured values of biomass-related variables for different cell types.

Cell type Cell di- λmeas CDM [g ⋅ L−1 ]/ OD[−]/CDM OD[−]/n Cells Remarks
ameter [nm] n Cells [g ⋅ L−1 ] [108 m ⋅ L−1 ]
[µm] [108 m ⋅ L−1 ]

Matured 7 750 4.26 3.32 14.1

yeast
Budding 7 750 3.49 3.20 11.2 Buds (30%)
yeast counted in
nCells
Chlorella 7.5 750 7.55 5.34 40.3 Organelles
vulgaris contribute
Bacillus 2*4 550 0.34 3.62 1.23
subtilis

evaluation:
I = I0 ⋅ exp (ε ⋅ c x ⋅ dpath ) (10.12)
The parameter ε is called the extinction coefficient. Historically the absorbance
A = − log(I/I0 ) is defined as a common logarithm and the optical density OD as:
A 1 I
OD = =− ⋅ log ( ) (10.13)
dpath dpath I0
Now some simplifications in this deduction have to be mentioned. The Lambert–Beer
law was originally stated for molecularly disperse systems at low concentrations. In
the case of measuring cells multiple scattering events occur making the real light ex-
tinction nonlinear with respect to light path length and biomass concentration. For
these reasons the light path length is standardized to 1 cm but not always in online
sensors where it is usually smaller. From Figure 10.10 it becomes obvious that also the
distance of the detector from the sample and the detector area have a direct impact
on the numerical results. In principle, during a measuring campaign always the same
device has to be used. A calibration curve has to be carefully determined with a set of
samples of known cell dry mass concentration.
The choice of the measuring wavelength has to ensure that no light absorption
occurs. For bacteria with characteristic gauges of 1 µm usually 600 nm is chosen. For
microalgae 750 nm is a good choice (no chlorophyll absorbance), and some online sen-
sors even measure in the near IR range at 880 nm. Eukaryotic cells with diameters of
several µm are considerably larger than the measuring wavelength. Here scattering at
the organelles or at surface structures will influence the results. This means also that
during a cultivation the relation between OD and CDM concentration will change fol-
lowing cell differentiation or adaptation. Table 10.3 shows some results for different
cell types.
The basic properties of cell suspensions measure different aspects of biomass and
can be relativized to each other. Optical density is considered to be linearly related to
 10.5 Measuring biomass – the great unknown ‘X’ | 265

cell dry mass concentration as well as cell number concentration. Small cells con-
tribute less to OD than the same number of larger cells (last column). However, the
higher weight of large cells compensates this effect so that the OD/CDM ratio is of-
ten in the same order of magnitude for different cells. The lower CDM/nCell values for
bacteria against yeasts correspond to their smaller volume. Cells are not simply par-
ticles but are semitransparent and contain other structures like organelles in the size
of the light wavelength. These contribute significantly to light scattering. As the size
and structural parameters (organelles) may change during a cultivation all relations
are subject to change as well.
While measuring OD off-line there is the possibility to dilute the sample until it
falls into the linear range of the calibration curve between CDM concentration and
OD; this is not the case for online turbidity sensors. Change of OD over two orders of
magnitude may happen during a cultivation. Some sensor systems measure transmis-
sion, side- and backscattering (reflection) for low, middle and high cell concentrations
to offer a signal that can be evaluated. Others can increase the sensing light intensity
according to the increasing biomass density. In every case careful calibration is nec-
essary.
The cell is even from a formal physical view more than a particle. It exhibits a
structure and a chemical composition, which may be exploited for measurement pur-
poses. From the principle of measuring cell size with a Coulter counter it is known
that cells do not conduct electrical current. The reason is that the cell membrane is a
chemical barrier against diffusion of ions and electrons, whereas the plasma is a good
conductor. A Nernst potential builds up over nearly all cell membranes and is con-
trolled by specific ion channels. Technically speaking the cell membrane behaves like
a battery or an electrical condenser. The electrical capacitance should be measurable
by its alternating current resistance, the impedance. A corresponding device is shown
in Figure 10.11. Via two electrodes an alternating electric field is applied leading, like
in capacitors, to a small charge displacement at the membrane. Here the cell behaves

AC generator
and planum charge displacement
electrodes
- - - -
+ + + +
- - - -
+ + + +

electric field lines

Fig. 10.11: Dielectric measurement of cells in a cultivation; only one cell is shown. Charge displace-
ment is indicated by + and -, while the membrane potential itself is not shown.
266 | 10 Measuring principles – how to put an end to the blind flight

light source detector

glass fiber
Fig. 10.12: Principles of optical sensing principles:
Fluorescence sensor to measure biomass; excitation
medium
light is provided by a LED or by an adjustable light
source delivering the whole spectrum of interest,
cells emission light spreads in all directions, then only
a small part reaches the glass fiber and finally the
detector (photodiode).

as an induced electrical dipole. Then the current can propagate inside the cells better
than in the medium. The impedance spectrum between 0 Hz and up to 10 MHz delivers
finally a signal depending on the fraction of cell volume in the medium and a shape
factor of the cells. For higher frequencies the organelles also contribute to the mea-
surement. The main fields of application are consequently large cells like mammalian
cells and yeasts as well as high cell density cultivations of bacteria. Dead cells (bro-
ken membrane) and other particles are not measured, which is why people speak of
‘viable cell density’ measurement.
Among the molecules of which cells consists two candidates are known to show
fluorescence: aromatic amino acids in proteins and NADH. Fluorescence is therefore
an option to measure biomass. This can be done using an in-line fluorescence probe
as shown in Figure 10.12 allowing us to record 2D fluorescence plots (Figure 10.13),
where emission is measured for a wavelength range of interest. As the protein content
may vary during a cultivation and the NADH/NAD relation depends on the cultivation
conditions as well, fluorescence measures only a facet of the cells depending on cell
composition and physiological state. Direct calibration is usually successful only in
well known processes, being a restriction for usage. An advantage is that conclusions
regarding the physiological state can be drawn.
Multivariate measurements promise to overcome limitations of single approaches.
However, this requires a lot of process knowledge and mathematical data evaluation.
In Figure 10.14 some different sensing signals are compared, all of them meant to mea-
sure biomass. Data are from a batch cultivation with a nitrogen limitation and subse-
quent accumulation of a nitrogen-free storage compound.
It is obvious that each of the different principles shows only one aspect of what
biomass is. In any case a combination of cell mass and physiological state is detected.
Furthermore, different physical sensing methods are corrupted by different distur-
bances, which have to be removed as good as possible by technical or data processing
means. Then the information has to be carefully interpreted. In this case cell wet mass
 10.5 Measuring biomass – the great unknown ‘X’ | 267

550 3063 -- 3500 [nm]

2625 -- 3063 [nm]
2188 -- 2625 [nm]
1750 -- 2188 [nm]
1313 -- 1750 [nm]
500 875.0 -- 1313 [nm]
437.5 -- 875.0 [nm]
0 -- 437.5 [nm]

flavin
Excitaon [nm]

450

400
vitamins and co-factors

350

300 aromac amino acids

350 400 450 500 550

Emission [nm]

Fig. 10.13: Typical outcome of a 2D fluorescence measurement, here for yeasts; the contribution of
protein and NADH florescence can be distinguished for further physiological interpretation.

5.0
Packed cell weight PCW *10
12 Fluorescence Fl & opcal density *10 OD
Dry cell weight DCW 4.5

4.0
10
Packed & dry cell weight [g.L-1]

3.5
8 3.0
Fl
2.5
6
OD 2.0

4 1.5

1.0
2
0.5

0 0.0
0 5 10 15 20
Culvaon me [h]

Fig. 10.14: Comparison of different online signals for biomass determination with an obvious devia-
tion during the cultivation.

and cell dry mass follow the growth process with a constant ratio of about ten, cor-
responding to a water content of the cells of 90%. In the late phase of the cultivation
cell dry mass overtakes cell wet mass due to accumulation of the storage compound.
Starch or PHB (polyhydroxybutyrate) granules contain only a little water. This inter-
268 | 10 Measuring principles – how to put an end to the blind flight

pretation is supported by the increase of optical density and the constant fluorescence
signal indicating constant protein content. Both optical measurements are very noisy
due to air bubbles and show a strong background signal from the medium as can be
seen by the high initial values.

10.6 Compiling a construction kit for sensors – a general

approach for biological sensors

All sensors have a general structure in common, as depicted in Figure 10.15. Firstly, the
analyte interacts with a ‘receptor’, which accepts only molecules with a characteristic
property. The receptor has to be as specific and selective for this property as possible.
Information flow is always coupled to an energy and/or mass flow. This can be diffu-
sion and/or chemical reaction energy. This primary energy flow to and in the recep-
tor – the primary measuring effect – is very small. Therefore, it has to be converted and
amplified by a second element of the sensor: the ‘transducer’. The transducer can be
unspecific because further energy or material transformation and amplification hap-
pens in the protected space of the sensor. In the transducer a much higher energy flow,
supported by auxiliary energy, is generated and modulated by the primary receptor
signal. Again, different kinds of physical or chemical effects can make up the specific
manifestation of the modulation. Finally, the transducer delivers the measuring signal
at a technically feasible energy level, commonly electrical energy. The signal is then
further amplified and processed for monitoring, control, and data storage.

enzyme
microorganisms substance electrode
analyte transistor
matrix cell receptors light signal
anbodies mass optode
nucleic acid piezoelectric

receptor transducer

Fig. 10.15: Basic structure in general of all sensors and here particularly for biosensors; the two
stages ‘receptor’ and ‘transducer’ are shown as rectangles, the corresponding material/energy
flows as arrows. The pink hash at the analyte/receptor interface indicates that only red diamonds
can bind but not the blue ones and no strings irrespective of the color. That underpins the concept
of selectivity.

Examples of this structure can be found among the sensors in the previous para-
graphs. For CO2 the characteristic property was high absorbance of IR. For oxygen
the semipermeable membrane is the receptor, while the active electrode is the trans-
ducer. In the case of pH the selective property of H+ ions is the charge, and the size of
the specifically permeable gel layer is the receptor.
10.6 Compiling a construction kit for sensors – a general approach for biological sensors | 269

The hardest problem in measuring biological molecules is to find specific proper-

ties and a selective receptor for this property. Thinking of enzymes for example, size or
surface charge are options. These parameters are used in chromatography. However,
the real thing to be exploited is the specific interaction of the enzyme with its sub-
strate. Otherwise it will not be possible to distinguish a particular enzyme from other
proteins during a cultivation. The great idea now is to employ biological principles to
solve technical problems. The receptor should include the chemical structure of a sub-
strate to measure an enzyme, in this way guaranteeing the most specific interaction.
This will work also vice versa. Furthermore, the receptor has to provide a signal to be
more or less nonspecifically further amplified by the transducer of course mounted
directly behind the receptor. The first sensor based on this principle was a glucose
sensor, where immobilized glucose oxidase is the receptor to react with glucose and
oxygen. Oxygen consumption is then detected by a classic Clark electrode as described
above acting as the transducer. In fact, two sensors are connected in series, where the
output of the first one is further detected by the receptor (oxygen permeable mem-
brane) of the second. This sensor, developed for use in medicine, was firstly described
by Clark and Lyons in 1962. Sensors including a biological component and typically
meant to measure biological analytes are called ‘biosensors’. Specific interactions be-
tween molecules are characteristic for living systems. These include enzymes and their
substrate or a substrate analog, antibodies and antigens, hormones, nucleic acids or
cell receptors. All these interactions are employed as receptors in biosensors.
Transducers themselves can be sensors e.g., for the product of the reaction be-
tween an analyte and an enzyme as receptor. This could be electrochemical cells to
detect e.g., pH. Optodes are employed in cases where the substrate or product is a
dye. This can be enforced by coupling a fluorophore to the reaction. The employment
of an ion specific field effect transistor (FET) is shown in Figure 10.16. Coupled with an
ion specific diffusion layer they are applied to measure ions in the solution. As many
such units can be placed on a sensor chip they can measure the nutrient composition
during the cultivation. Cross-sensitivities are actively compensated by the different ion
specific layers on the chip. When used as biosensors for organic compounds they are
covered with an additional biologically active layer to transform the organic analyte to
a measurable ion. Other possibilities are under investigation or in use. One example is
the employment of piezoelectric crystals. The additional mass induced by binding of
the analyte to a ligand changes the resonance frequency of the piezoelectric crystal.
A last example for a biosensor in this chapter is shown in Figure 10.17. Here mi-
croorganisms have the job of measuring heavy metals in wastewater. The kernel of
the idea is a recombinant promotor, which is inducible by heavy metals. Coupled to
the promoter is the LacZ gene from E. coli enabling the microorganisms to produce
galactosidase and therefore to metabolize lactose. Applying a wastewater sample to-
gether with lactose leads to cleavage of the lactose and respiration on glucose and
galactose, but only if the wastewater contains heavy metals. The corresponding oxy-
gen consumption is measured with a classic pO2 electrode. Running this electrode re-
270 | 10 Measuring principles – how to put an end to the blind flight

bias voltage
reference enzyme layer
ion selecve layer
p-silicon electrode
housing
amino acid

NH4+
source drain

source-drain-voltage source-drain-current

Fig. 10.16: Biosensor based on an enzymatic reaction and a FET. An amino acid ammonia lyase
cleaves a dedicated amino acid. The charge of the produced ammonium ions is strong enough to
act as input for the ion sensitive gate, modulating the source to drain current being the output sig-
nal. The bias voltage has to be applied to adjust the working point.

- +

rubber cup

cathode anode

Teflon membrane
immobilized
microorganisms

dialysis membrane

Fig. 10.17: Scheme of a biosensor with immobilized living microorganisms to measure heavy metals
in wastewater.

quires control of the different measuring and flushing phases during a measurement
cycle and evaluation of the dynamic responses.
Biosensors are for many measuring tasks the only possibility for online sensing.
Despite this potential there are some drawbacks to be mentioned, which include lack
of long term stability or high temperature resistance during sterilization. Solutions to
 10.7 Questions and suggestions | 271

such problems can be found in avoiding direct contact with the cultivation suspen-
sion by a separating membrane or by mounting the sensor in a bypass as in quasion-
line devices. Typical fields of application are environmental biotechnology or food
technology e.g., as single use disposable sensors. Modern applications of biosensor
principles also target measuring in microtiter plates or even in the intracellular space.
In fact, sensory organs in animals work also in accordance with this principle. Light
for example triggers a photochemical reaction of rhodopsin – being the receptor – in
the retina, then an ion channel opens triggering in the nerve membrane – being the
transducer – an electric impulse, where the membrane charge is the auxiliary energy
transducer.

10.7 Questions and suggestions

1. Which effects lead to changes of the weight of a reactor during cultivation?

2. The λ sensor built in the exhaust pipes of cars measures basically oxygen
content of the exhaust gas. Find out what is the specific element of this sensor.
3. Calculate the voltage that is generated at a silver electrode to measure Ag+
ions. Find the values of the natural constants in the appendix. T = 293 K;
charge number (valence) z = +1; φ∘ = 0.8 V (from a table book); γ = 1 (as-
sumed); aAg+ = nAg+ ⋅ 1 mol−1 ⋅ kg.

1. Evaporation, dosing of acid and base, sampling, or oxidation of sugar by the mi-
croorganisms, where CO2 in the exhaust gas has a higher molecular mass than O2
in the fresh gas.
2. Diffusion of O−2 through a crystal of yttrium – stabilized zirconium dioxide (ZrO2 ).
3. nAg+ = 1mol
kg : ϕ meß = 0.8 V + ln(1) ⋅ 0.0257 V = 0.8 V Change of cation concentra-
tion by a factor of ten results in 59.2 mV measuring voltage (ln(0.1) = −2.3)
11 The practice of fermentation – a step by step
guide through the workflow
In the previous chapters the processes inside a reactor were highlighted and we de-
scribed how they can be influenced by the inputs provided by the reactor. Neverthe-
less, it is another problem to stand in front of a real bioreactor and to consider what
to do next. In this chapter, a step by step guide for the cultivation of microorganisms
is given and broken down into single actions. The main aim is to become acquainted
with fermentation systems and learn how to solve technical problems using concrete
examples. Since the single steps for conducting fermentations in a bioreactor are in
general quite similar, we encourage transferring the presented steps to other cultiva-
tions. By following the instructions of this guide, even inexperienced users should be
able to run a standard bioprocess. The checklists and example protocols given in the
particular paragraphs can easily be transferred to other fermentation systems or to
different kinds of cultivations.

11.1 Structuring the reactor and the cultivation – zooming in from

principle understanding to visibility of process details

Before beginning with the actual work, it is mandatory to comprehend the mate-
rial flows in the reactor system and to identify the respective technical realization.
A graphical representation is given in Figure 11.1, showing basic reactor modules.
This hand drawing may represent the impression of somebody having the first look
onto an unknown bioreactor. In the next step it is important to achieve a structured
mind mapping regarding inputs and outputs of gas, steam, and liquid streams and
why they are required in the respective sections of the reactor. Follow all necessary
media (gas, fluids) along the material flows! Standard procedures for how to set up the
reactor vessel and how to prepare the preculture are also mandatory. Therefore, this
guide is split into four parts: cultivation design, cultivation preparation, the cultiva-
tion itself, and data evaluation, providing concrete and detailed solutions concerning
the exemplary fermentation process. In the following paragraphs we will move closer
towards an engineering view, to become more familiar with further details and the
relevance of single reactor parts. Handling of the reactor and timely structuring of
the fermentation become clear by performing a real concrete fermentation. In this
chapter an aerobic fed-batch process is chosen, here with Bacillus amyloliquefaciens
producing phytase, an extracellular phosphatase, which is excreted into the cultiva-
tion suspension.

https://doi.org/10.1515/9783110315394-011
 11.2 The reactor – a complex assembly of parts and modules | 273

Fig. 11.1: A graphical sketch of a pilot reactor. Basic modules are

visible; more details only open up by getting closer and running
hands-on cultivations (© bioengineering).

11.2 The reactor – a complex assembly of parts and modules

The cultivation will be performed in pilot scale using a stirred tank reactor (STR). This
is the most commonly used reactor type for bioprocesses. Approaching the reactor, the
first thing we will see is a stainless steel cylinder with a lot of periphery attached next
to a rack made out of more steel with little displays (Figure 11.2). For clarification each
single element will be identified and classified regarding material and energy flows.
Every organism needs nutrients. In the case of bacteria, these are oxygen for cell
respiration and culture media, which are designed to provide all elements the cells
need. To supply each single cell with enough nutrients it is necessary to mix the cul-
ture broth and to disperse the gas bubbles. In stirred tank fermenters this is accom-
plished by mechanical agitation. The pressurized inlet gas enters the reactor vessel
from the bottom. The inlet gas, usually air owing to its low cost, provides oxygen for
aerobic cells during cultivation and removes gaseous byproducts. A reactor jacket al-
lows heating and cooling of the culture broth. As a training example, a commercial
pilot reactor with a total volume of 19 L and a work volume of 12.6 L is used. Table 11.1
gives an overview of the technical specifications. Such a table is usually given in the
manual or has to be set up by the user.
The vessel itself consists of stainless and acid resistant steel. It is fixed with clamps
on a rack. The five lateral ports in the bottom of the vessel allow the arranging of sen-
sors in the fermentation broth. During fermentation, the broth can be observed via a
longitudinal window of security glass. With an operating temperature of 150 °C and
274 | 11 The practice of fermentation – a step by step guide through the workflow

condenser exhaust gas filter

gas flow outlet
feed
base / acid
gas flow inlet
air filter an foam

foam
sensor

heang /
cooling water
tempering
jacket baffle

srrer measurement
heang /
cooling water cabinet
sparger

pH sampling
sensor point
pO2
sensor motor temperature sensor

Fig. 11.2: Scheme of the pilot scale ‘fermenter’ for use in development labs; the symbols represent
parts and devices, which have to be identified at the real reactor.

Tab. 11.1: Specifications of the bioreactor system.

Component Technical data

Vessel Stainless steel 316 L
Heating circuit Electrical heater; heat exchanger for cooling circuit
Mechanical seal Silicon carbide; glycerol lubricated
Stirrer Blade stirrer, baffles
Inlet air Absolute filter; gas mixing station
Agitation Belt driven from bottom; double mechanical seal
Aeration Submerse; stainless steel ring sparger

an operating pressure of 2.5 bar a sterilization by superheated steam is possible. The

lid is fixed by tightening wingnuts on studs. It is sealed by an O ring seal made of a
synthetic elastomer rubber (EPDM). Like the vessel, the lid is made of stainless and
acid resistant steel. The ports in the lid are round threads. Thus they are hygienic and
mechanically stable. Gas inlet and gas outlet are connected to the reactor via the lid.
In our setup an autosterile filter is attached to the gas flow inlet. Medium, base, acid
and antifoam feed as well as the foam sensor are connected to the lid. The stirrer is
belt driven by an electrical engine (AC; 1.1 kW; max 1500 min−1 ) from the bottom. If
 11.3 Structuring the cultivation process – from planning to final data acquisition | 275

necessary, a gear reduction up to 1:5 is possible. The power electronics for the engine
is located in a separate cabinet.
To prevent leakage between shaft and vessel a double mechanical seal is used as
shaft bearing (see Chapter 5). A foam separator is integrated to eliminate foam and
remove particles from the gas discharge. The resulting pressure drop between vessel
and gas discharge enables the foam to be pushed against the rotor. The mechanical
forces induced by the rotor separate the gases, which then can pass off via the de-
gassing chamber and the exhaust gas line. The liquid particles are thrown back into
the foam mass.
Process parameters are controlled by 19” modules integrated in a measurement
cabinet and a higher level process control based on a graphically programmable pro-
cess control system. Temperature regulation is necessary to change and control the
heating of the bioreactor during the autoclaving process and to keep the temperature
constant during the cultivation process. A Pt-100 temperature sensor is used to control
the cultivation temperature. The cultivation broth is tempered by a double wall vessel.
An electrical heater, directly connected to the double wall vessel, sets the temperature
in the primary circuit. The electrical heater is connected to a heat exchanger, which is
connected to an external cooling circuit.

11.3 Structuring the cultivation process – from planning to final

data acquisition

This subsection presents a typical cultivation, an aerobic fed-batch cultivation of

Bacillus amyloliquefaciens. A systematic approach along the sequence of necessary
steps is given to explain the typical procedure regarding planning, assembling, and
commissioning of a standard fermentation. We take it for granted that the main con-
nections for gas supply and cooling water are already available. Air is usually supplied
by a compressor and should be free of oil and moisture. Pneumatic valves should be
utilized to provide the process gas at 1–3 bar. The water used should be of a quality
that meets the process requirements. We assume furthermore that all analytical de-
vices and methods as well as the needed software, which may be necessary for sample
processing and the evaluation of process data, are available and were tested before.
The utilized process control system has to be attuned for the cultivation purpose by
either self-programming or consultation with a software manufacturer. In the case of
a fed-batch process, feed controllers have to be implemented and parameterized in
order to control the feed pump flow rate.
Before starting to work with the reactor, some safety instructions have to be inter-
nalized:
– Never heat an empty reactor. Devices within the fermentation vessel could be de-
stroyed.
276 | 11 The practice of fermentation – a step by step guide through the workflow

– Always use a safety overpressure valve to prevent a pressure increase in the case
of a blocked output filter. In the case of exceeding the maximum pressure, the
security glass inside the fermenter vessel must burst.
– Make sure to use constant gas pressure < 0.2 MPa (2 bar). A security gas valve can
reduce pressure fluctuations.
– During sterilization, the exhaust gas line of the reactor system has to be open, to
compensate possible overpressure during the sterilization process.
– Sterilize dangerous and corrosive liquids and storage bottles separately. Transfer
the chemicals to the sterile storage bottles under sterile conditions.
– Wear goggles and gloves and follow the safety instructions while handling corro-
sives.

Prior to realizing the experimental setup, it is important to carefully consider the pro-
cess strategy and data evaluation. Figure 11.3 shows a principle pattern of an exper-
imental fermentation procedure. Summarized under the term ‘upstream processing’
are the preparation of medium, the sterilization of reactor components, and the in-
oculum preparation.
At first, it is necessary to prepare culture medium (red box) in sufficient quanti-
ties. Especially for fed-batch processes, calculation of the required amount is neces-
sary since culture medium is needed in sufficient amounts for preculture, batch phase,
and fed-batch phase. The second work package (brown boxes) deals with preparing
the inoculum from strain maintenance in the lab up to a sufficient amount of precul-
ture for inoculation of the reactor. Depending on the strain, medium, and the objective

assembly of media strain

reactor system preparaon maintenance

filling the preculture

reactor vessel inoculaon

autoclaving the preculture

reactor preparaon

main culture
inoculaon

fermentaon sampling, reassembling,

process analysing cleaning

Fig. 11.3: Block diagram of an experimental fermentation procedure; the red, brown, gray, and cyan
boxes mark the steps in the workflow. The colors represent related steps belonging to a respective
work package.
 11.4 Media preparation | 277

of the experiment the amount of inoculum needed can vary. However, the inoculum
volume used for this experiment was 5% in terms of the liquid reactor volume. In-
oculation needs precise timing as a vital culture in the exponential growth phase is
required. So the lab part of this work package has to start first. Media preparation and
preculture preparation are the typical upstream parts of the process. In the third work
package (gray boxes) the bioreactor is set up by assembling the fermentation vessel
and peripheral devices. Filling with medium and sterilization are the next steps. Inoc-
ulation finally completes this work package. The cultivation itself (cyan boxes) starts
with transferring the axenic preculture into the sterilized reactor and activating the
process control system, which controls the feeding rate and records all digital process
parameters. In the next four subsections the working lists according to the four work
packages are further elaborated.

11.4 Media preparation

The media composition used for the preculture and the main culture of strain Bacil-
lus amyloliquefaciens are given in Tables 11.2 and 11.3–11.6 respectively. An amount of
10 mL/L trace elements solution has to be added to both solutions. The composition
of the trace element solution is given in Table 11.4 while the stock solution phosphate
and glucose, which are used for the fed-batch process in the main culture, are listed
in Tables 11.5 and 11.6. For the purpose of the exemplary cultivation a modified syn-
thetic medium was used; see also Section 3.3. Usually, the medium is sterilized in the
preculture flask or in the fermentation vessel. However, a sterilization in advance of
its use is also possible. The most popular methods for media sterilization are filtration
or thermal treatment.

Tab. 11.2: Components of the preculture medium; pH has to be adjusted to 7.5 and sterile glucose
solution should be added under sterile conditions after autoclaving; further details are given in the
Section 3.3.

Components Concentration [g/L]

(NH4 )2 SO4 2.6
MgSO4 ⋅ 7H2 O 0.41
HEPES 12
Trisodium citrate dihydrate 1.14
KH2 PO4 3.3
Glucose monohydrate 30
278 | 11 The practice of fermentation – a step by step guide through the workflow

Tab. 11.3: Components of the main culture medium. Potassium dihydrogen phosphate and glucose
should be added after the autoclavation process under sterile conditions using feed bottles.

Components Concentration [g/L]

(NH4 )2 SO4 2.6
MgSO⋅4 7H2 O 0.41
Trisodium citrate dihydrate 1.14
KH2 PO4 Potassium dihydrogen phosphate 0.3
Glucose monohydrate 42

Tab. 11.4: Trace elements solution (100 fold).

Components Concentration [g/L]

CaCl2 0.291
Na2 MoO4 0.024
FeSO4 ⋅ 7H2 O 0.52
ZnSO4 ⋅ 7H2 O 0.72
MnCl2 ⋅ 4H2 O 1.00
CuCl2 ⋅ 2H2 O 0.085

Tab. 11.5: Components of the phosphate feed solution.

Components Concentration [g/L]

KH2 PO4 5.0

Tab. 11.6: Components of the glucose feed solution.

Components Concentration [g/L]

Glucose monohydrate 600

11.5 Preculture preparation

Preculture preparation has to provide a sufficient amount of active cells to inoculate

the bioreactor and is generally accomplished under sterile conditions, e.g., in a biolog-
ical safety cabinet. Different cell banking systems are used to store the desired strain,
e.g., in liquid nitrogen (long term storage), agar plates, or liquid cultures (mid and
short term storage). In the regarded case, the preculture is present as an agar plate
culture. To get liquid precultures of Bacillus amyloliquefaciens sterile 250 mL Erlen-
meyer flasks containing 20 mL sterile preculture medium (Table 11.2) are prepared.
Via a flamed inoculation loop bacterial cell mass is transferred from the agar plate
to the Erlenmeyer flasks. After sealing, the liquid precultures are incubated at 37 °C
and rotated with 150 rpm on an orbital shaker. As soon as the liquid cultures enter the
 11.6 Reactor installation and setup | 279

Tab. 11.7: Checklist for preculture preparation.

Tasks for preculture preparation Done

a) Prepare preculture medium for liquid shaking flask cultures

b) Prepare Erlenmeyer flasks (250 mL) with sterile closures for gas exchange
c) Fill each flask with 20 mL medium and sterilize these for 20 min at 121 °C
d) Inoculate liquid media with preculture from agar plate under sterile conditions
e) Incubate liquid precultures on an orbital shaker at 37 °C and 150 rpm
f) Start the preparation of another preculture generation in advance before the
first cultures enter the stationary phase
g) Autoclave older precultures if they are no longer required

exponential growth phase, it is recommended to prepare from this charge a second

charge of liquid precultures to ensure a proper adaption of the microorganisms to the
new environmental conditions. The second preculture charge can then be used as in-
oculum for the main cultivation process. It is essential to prepare sufficient inoculum
with the appropriate cell density at the end of the exponential growth phase. Growth
conditions should enable the generation of high cell densities within a short period.
Preliminary examination of preculture growth rates is therefore helpful to determine
an appropriate inoculation time, which should be aligned with the main cultivation.
Table 11.7 lists the single steps to take into account for the preparation of precultures.

11.6 Reactor installation and setup

All major bioreactor components have to be set up to operate the fermentation pro-
cess. Medium has to be prepared in advance (see Figure 11.4). Several openings in the
reactor cover and the reactor hull allow the installation of the periphery. It is important
to make sure that the sterile air inlet device is installed correctly (connection defined
by tubular die) before the reactor cover devices are installed. Furthermore, it has to be
taken into account that the ports for inoculation and sample taking are easily acces-
sible. Septa have to be inserted in the proper direction. Do not connect the external
flask to the reactor prior to autoclaving. To verify the functionality of the setup after
assembling, it is mandatory to track the complete pathway of the gas and the cooling
water flow. Online sensors are connected to the system via the control unit. In a typical
fed-batch cultivation, the necessary installation and setup steps will be described in
detail in the following sections.
280 | 11 The practice of fermentation – a step by step guide through the workflow

6
2

7
Fig. 11.4: Picture of the example reactor; the main
parts are indicated by numbers: (1) exhaust gas
cooling and filtration; (2) autosterile filter with
air inlet, ports for metering, illumination, an-
tifoam sensor, exhaust gas port, and manometer;
3 (3) lateral ports for pH, pO2, pCO2, temperature,
and sample port; (4) big sample port; (5) motor;
5
(6) pumps for acid/base/antifoam metering and
4 feed; and (7) measurement cabinet (© bioengi-
neering).

11.6.1 Fermentation vessel setup

For the setup of the fermentation vessel, the following parts have to be mounted:
sparger, air input including sterile filter for gas inlet, agitator shaft including six-blade
Rushton stirrer, baffle, dummy plugs, septa, and optionally an illumination device to
illuminate the inward reactor vessel. The sparger is needed to disperse oxygen in the
liquid phase. It is usually installed below the impellors. Before entering the bioreactor
sparge line, the airflow passes a sterile filter (pore size 0.2 µm) to minimize the risk of
contamination. The supplied air is then released into the liquid phase as gas bubbles
and dispersed by the agitation and baffling system. It is necessary to verify the func-
tionality of the whole gas pipe system before the reactor setup is continued. After the
setup of the sparger, the tubing for gas inlet and outlet including sterile filters, and
the agitation and the baffling system are mounted. The fermentation vessel has to be
hermetically sealed when it is connected to the gas line. By gassing the sealed reac-
tor system with air, it is possible to verify the functionality of the gas pipe system by
routing the gas flow via a bypass into a water bottle and observing the bubbling.
 11.6 Reactor installation and setup | 281

11.6.2 Peripheral devices

The material flows can be driven by pressure or pumps and have to be directed by
and through additional peripheral devices. Some of these are connected to the inner
reactor space via the reactor head; see Figure 11.5.

4 5

2
6
8

1 7 10

Fig. 11.5: Head of the reactor; the parts to be mounted are indicated by numbers: (1) condenser;
(2) valve; (3) autosterile filter; (4) illumination device; (5) antifoam sensor; (6) acid/base/antifoam
metering; (7) manometer; (8) feed; (9) exhaust gas outlet; and (10) septum for inoculum (© bioengi-
neering).

Overpressure security valve: Mount an overpressure relief valve for transient overpres-
sure protection during cultivation or sterilization. Additionally, the installation of a
manometer is recommended to monitor pressure history during the autoclaving pro-
cess.
Sample unit and fittings: Installation of a sterile sampling unit and fittings for
the addition of inoculum, medium, or stock solution feed as well as solutions for pH
adjustment (acid/base) are mandatory. Liquid compounds are pumped into the fer-
mentation vessel from sterile storage bottles.
Outgas condenser: To prevent condensation of water and the related risk of a
blocked gas flow and to protect the off-gas analyzer, the installation of an outgas con-
denser is necessary. The condensed water will flow back into the reactor vessel instead
of clogging the outgas filter. It is common practice to use an intermediary bottle be-
tween the outgas condenser and the output gas filter. To improve the protection of the
output gas filter against clogging, this bottle can contain a small amount of antifoam.
Gas outlet filter: To guarantee the preservation of a sterile reactor system after
autoclaving a gas outlet filter is needed. For this purpose, an autoclavable sterile gas
filter with a pore size of 0.2 µm is used. To avoid clogging it is mandatory to ensure that
the gas outlet filter stays dry at all times.
282 | 11 The practice of fermentation – a step by step guide through the workflow

Antifoam sensor: Mount the antifoam sensor and the antifoam stock bottle. The
antifoam sensor acts as a level probe. As soon as foam reaches the sensor, the sensor’s
electrical conductivity changes, which leads to the activation of the antifoam pump.
During the cultivation process the formation of foam due to aeration, agitation, and
the presence of foam producing or foam stabilizing substances like proteins, polysac-
charides, or fatty acids can occur. Foaming can lead to a blockage of the outlet filter
and gas lines and has to be avoided. Antifoam agents, mechanical foam separators, or
ultrasonic treatment can be used to reduce foam formation. However, surface tension
lowering antifoams are usually utilized to reduce energy consumption.
Peristaltic pumps: To achieve precise and reproducible media feeding and meter-
ing of acid, base, and antifoam peristaltic pumps are used. These types of pumps are
easy to handle and maintain. Via the dedicated control buttons, the rotation speed of
a peristaltic pump can be regulated. It is noteworthy that the volumetric flow rate of a
peristaltic pump depends on the internal diameter of the tubing. In preliminary mea-
surements the correlation between the rotation speed and the respective tubing has to
be identified. Thereby a calibration of the favored flow rate range is possible. The re-
quired time for pumping a certain liquid volume into the reactor, e.g., 10 mL, has to be
measured. Dividing the pumped liquid volume by the required time gives the desired
flow rate (mL/min). The bioengineering base control unit offers several pump support
brackets to place the peristaltic pumps, so valuable bench space is saved.
Mass flow controllers: To supply the microorganisms with a defined volumetric
gas flow, mass flow controllers are installed. Use gas tubing to channel the inlet gas
between the gas lines and mass flow controllers inside the fermentation vessel. It is
common to utilize a mixture of air and O2 in aerobic fermentation processes. Airflow
calibrators are recommended to enable an accurate and consistent measurement of
the desired flow rate. Sterile filters (pore size 0.2 µm) are integrated to sterilize the
gases.
As soon as the setup is completed, the functionality of the peristaltic pumps, mass
flow controllers, gas inlet, sensors, and the computer connection has to be checked
once more. If all devices work properly, the sterilization process can be launched.

11.6.3 Sensor installation

During cultivation pH, dissolved oxygen (pO2 ), and temperature in the fermentation
broth have to be measured and controlled.
pH sensor: Calibration has to be performed before autoclaving. The calibration
buffers have to be selected according to the pH range of the fermentation. For standard
fermentations a pH 4.0 buffer and a pH 7.0 buffer are recommended. The calibration
of the pH sensor has a two point gauging system.
pO2 sensor: A two point gauging calibration is made after autoclaving while stir-
ring. Zero point of calibration is N2 saturated medium, while the slope calibration
 11.6 Reactor installation and setup | 283

Fig. 11.6: Example pictures of an exemplary process control system ‘BioProCon 2016’, a graphical
process control software that was developed at the Karlsruhe Institute of Technology (KIT), Institute
of Bioprocess Engineering (© KIT).

point is determined in an air saturated medium. Although pO2 sensors usually com-
pensate temperature effects automatically, it is preferable to calibrate the sensor at the
fermentation temperature since the saturating oxygen concentration varies with tem-
perature, pressure, and concentration of dissolved media substances. The maximal
dissolved oxygen saturation concentration at 37 °C in water is 6.72 mg O2 /L.
Temperature sensor: A two point gauging system is used to calibrate this sensor
type. A temperature range similar to the fermentation process is recommended, e.g.,
37 ∘ C for cultivation and 121 ∘ C for autoclaving. Heated water can be used for the cali-
bration process. Please have safety arrangements for hot boiling liquids in mind.
The process control system is necessary to enable a fully automatized monitoring
of the cultivation as well as the storage of all recorded online data. For this fermenta-
tion, a graphical software based process control system is used. Monitoring is possible
via inbuilt displays; Figure 11.6. All sensors used to control and monitor the process
need to communicate with the process control system. This is usually ensured by a
supplier designed communication bus system via analog or digital input/output sig-
nals.

11.6.4 Autoclaving the bioreactor

To ensure sterile working conditions, the fermentation vessel and all devices that will
be in direct contact with the medium have to be autoclaved. Before starting the steril-
ization process, several specific preparations have to be made. First, the fermentation
284 | 11 The practice of fermentation – a step by step guide through the workflow

vessel has to be filled with cultivation medium. Therefore, all unused ports must be
closed with septa or screw caps. The liquid working volume of a bioreactor is usually
70 or 90% of its total volume to keep some headspace for foam formation; the volume
increases by aeration and substrate feeding. In fact, the bioreactor for our example
cultivation is only filled up to 50% at the beginning (batch phase) to leave sufficient
space for the feed solutions during the fed-batch phase. With exception of the pO2 sen-
sor, each sensor has to be calibrated and the sensor connectors have to be protected
by a cap. Each sensor should be disconnected from its cable during the sterilization
process to avoid damage to the sensor cable. It is important to check once more if the
overpressure security valve is installed in the reactor vessel. The gas outlet has to be
closed at the beginning of autoclaving and will be opened later for pressure compensa-
tion during the autoclaving procedure. Other tubes that are in contact with the liquid
medium have to be closed. All storage bottles and pump lines have to be sterilized
separately. Nondangerous liquids can be filled into the storage bottles and autoclaved
directly. Acid, base, or liquid substances with high evaporation losses should be au-
toclaved separately. The utilization of pressure resistant bottles is recommended. The
attached storage bottles with pump lines can be autoclaved with some drops of water
in them. After the sterilization, the sterile acid, base, or liquid substances with high
evaporation losses are filled into the sterile storage bottles. Please make sure to use a
ventilation line with a sterile filter, pore size 0.2 µm, connected to each storage bottle
to allow pressure compensation. The liquid level in the bottles, which shall be steril-
ized in an autoclave, should not exceed two-thirds of the maximum bottle volume. A
checklist to get ready for the autoclaving process is given in Table 11.8.
In our setup, an autosterile ceramic filter for manual sterilization is used. One
main advantage of this system is the possibility of in place sterilization along with the
reactor autoclavation process. The steam created in the vessel passes the opening in
the filter housing, sterilizing the filter and the inlet piping. During fermentation, the
filter housing is lifted to a position where the steam inlet is sealed by the reactor lid.
However, other reactor systems often use other sterile filters instead of an autosterile
filter for manual in place sterilization. In that case, the gas inlet filter has to be auto-

Tab. 11.8: Checklist to prepare the autoclaving process.

Tasks before autoclaving Done

h) Calibrate the pH sensor

i) Check the functionality and precalibration of other sensors
j) Mount the reactor vessel including the overpressure valve
k) Prepare medium for batch phase and fill the reactor vessel
l) Disconnect all cables and protect the cable connectors with protection caps
m) Mount an autoclavable gas inlet and gas outlet filter, pore size 0.2 µm
n) Prepare storage bottles and tubing
 11.6 Reactor installation and setup | 285

claved or sterilized externally prior to autoclaving the reactor system. The installation
of sterile filters has to be considered carefully for each reactor system individually.
Oil- and moisture-free air at a pressure of 1–3 bar is led into the medium via the
sterile filter and the submerse stainless steel ring sparger; see also Section 11.6.1. Ex-
haust gas will pass through and exit the vessel. The exhaust gas will be moisturized
by passing through the liquid phase. To prevent evaporation and blocking of the off-
gas filter, water in the exhaust gas is condensed in the exhaust gas condenser and
channeled away towards the vessel while the gas can escape through the exhaust gas
pipe.
Utmost attention is necessary during autoclaving, since all metal parts will get
quite hot and the fermentation vessel will be under pressure. Safety glasses and ther-
mal gloves are mandatory. During the sterilization procedure, different stages are
passed through. Figure 11.7 shows the piping and instrumentation. The liquid reactor
vessel content is heated with steam via the reactor jacket. The sterilization tempera-
ture has to be maintained constant for a certain period before the reactor vessel can be
cooled down with water to cultivation temperature. To facilitate correct autoclaving,
a proper heat exchanger network is mandatory.
At first, the temperature of the bioreactor system is increased from room temper-
ature to 95 °C in the heating-up cycle. During this period the autosterile filter for gas

exhaust gas

air 2-7 bar

M

4,5 kW heater

waste water (pressureless) PC

cooling water 3-4 bar(g)

cooling water 2-4 bar(g)
cooling water return

Fig. 11.7: Piping and instrumentation diagram to check the heating and cooling circuits.
286 | 11 The practice of fermentation – a step by step guide through the workflow

inlet is closed and the condensate drainage of the outgas condenser is open (see Fig-
ure 11.7). (In the case of other sterile filter techniques, the sterile filter is autoclaved ex-
ternally). The air supply has to be closed during the whole heating-up process. When
the temperature has reached 95 °C, the sterilization cycle for the gas outlet section
is induced by closing the gas outlet valve to a minimum. Thereby a minimal flow of
hot steam passes through the gas outlet section, ensuring a proper sterilization. The
sterilization cycle is induced by increasing the reactor temperature to 121 °C. This tem-
perature has to be kept for at least 20 min. According to vapor pressure of water an
overpressure of 1 bar (2 bar absolute, vapor pressure of water at 121 °C) will build up
in the reactor vessel at this temperature. However, due to the opened gas outlet valve,
the overpressure is adjusted to 1 bar during the autoclavation process. The steriliza-
tion cycle is terminated by reducing the temperature to 100 °C. Once 100 °C is reached,
the following steps have to be conducted: carefully open the gas outlet valve, activate
the exhaust air cooling system, activate sterile filter for gas inlet, open the gas sup-
ply, and close the condensate drainage of the outgas condenser (for autosterile filter
only). Due to this procedure, the generation of a vacuum inside the fermentation ves-
sel is prevented. By cooling down the reactor system to cultivation temperature (37 °C),
autoclaving is finished. Each peripheral unit, like storage bottles, tubing, and sensor
cables, can be connected to the system as soon as the reactor vessel has a tempera-
ture of 75 °C or less. Table 11.9 gives a summary of the single steps of the autoclaving
process in the form of a checklist.

Tab. 11.9: Checklist to perform the autoclavation process.

Tasks for autoclaving Done

a) Close the exhaust gas cooling system, open the gas outlet section
b) Check if the air supply is closed, lower the sterile filter for gas outlet, open the
condensate drainage of the outgas condenser
c) Raise the temperature to 95 °C
d) When 95 °C is reached, close the gas outlet valve and the condensate drainage
valve to a minimum
e) Raise the temperature to 121 °C
f) When 121 °C is reached, keep this temperature for at least 20 min
g) Lower the temperature to 100 °C
h) When 100 °C is reached, lift sterile filter for gas outlet, close the condensate
drainage valve, open the gas supply, and open the gas outlet valve
i) Lower the temperature to 75 °C
j) Connect all peripheral units: storage bottles, tubing, and sensor cables
k) Adjust cultivation temperature (37 °C) and calibrate the pO2 sensor
 11.7 Managing the actual cultivation process | 287

11.7 Managing the actual cultivation process

The fed-batch cultivation process is started by inserting the preculture into the fermen-
tation vessel. Before inoculation, the agitator speed, gas flow rate, and the pH have to
be set to the requirements of the experiment. All devices for measurement and control
have to be connected to the system and functional. If not performed already, the pO2
sensor has to be calibrated. Table 11.10 gives a checklist summarizing the tasks before
inoculation.

Tab. 11.10: Checklist before inoculation.

Tasks before inoculation Done

a) Reactor system at cultivation temperature

b) Connect each peripheral device to the system if not already done, including
storage bottles, tubing, and sensor cables
c) Calibration of the pO2 sensor
d) Adjust set-points of cultivation parameters (agitator speed, gas flow rate, pH,
...) to the growth specification of the start conditions in batch mode
e) Inoculum prepared and inoculation bottle connected
f) Process control system (PCS): data recording and controller activated
g) Inoculation of preculture

Immediately after preculture inoculation, the first sample should be taken. To ensure
a sterile sampling procedure the tubing of the sample units has to be rinsed before.
The washing procedure is similar to the sampling procedure. The forerun cultivation
broth can be discarded. An appropriate sampling cycle is recommended regarding the
growth rate of the microorganism. For Bacillus amyloliquefaciens a sampling cycle of
every one to two hours is sufficient. During cultivation an increased foam formation
can occur, therefore it is necessary to use an antifoam detergent or a mechanical foam
separator. Both chemical and mechanical foam reducers can be controlled by imple-
menting a level probe.
For the presented aerobic fed-batch cultivation of Bacillus amyloliquefaciens a two
stage process is used to produce phytase. In the first stage, the batch phase, biomass
is produced until the nutrient phosphate is almost depleted. In the second stage, the
fed-batch phase, product formation is carried on by continuously feeding phosphate.
Since the synthesis of phytase is enhanced under phosphate starvation, the fed-batch
phase is necessary to guarantee an ongoing phytase production. An additional glucose
feed is used to supply the bacteria with a sufficient amount of carbon and energy to
maintain metabolism.
Stage 1 – Batch: After a short lag phase, the microorganisms enter the exponential
growth phase. By depletion of the phosphate source, the metabolic activity of the cells
288 | 11 The practice of fermentation – a step by step guide through the workflow

is decreasing. An increasing pO2 value is a sign of the reduced metabolic activity. At

this point, the process is switched to fed-batch mode.
Stage 2 – Fed batch: During fed-batch mode the culture is supplied with glucose
and phosphate. The feed rate of the phosphate source is adjusted to the target growth
rate (μ). By the marginal supply of phosphate the production of phytase is induced.
The concept is given in Chapter 7 on fed-batch processes.
The glucose feed strategy is to adjust a continuous volumetric feed rate to ensure
a sufficient, nonlimiting glucose concentration during fed-batch mode. An accumu-
lation of glucose should be avoided and thus the supplied amount of glucose should
be as high as the glucose consumption of the microorganisms. Furthermore, the feed-
ing profile of glucose is related to the phosphate feeding profile and can be calculated
by the related yield coefficient of the microorganisms for glucose and phosphate. By
analyzing the biomass, the glucose and the phosphate concentration during fermenta-
tion, these yields can be determined (Y X,glucose = g cell dry mass/g glucose; Y X,PO4 = g
cell dry mass/g phosphate). Table 11.11 and Table 11.12 give an overview of the cultiva-
tion and feed parameters of the fermentation process.
Process parameters have to be ensured to be constant during the whole pro-
cess time. The actual values for temperature, agitator speed, flow rate, and pH of
the medium have to be checked continuously. Additionally, qualitative observations
like color, turbidity and smell of the cultivation broth, cell morphology, or possible
contaminations have to be protocolled. For the determination of more cultivation
parameters, samples are taken in defined sampling cycles via the sampling device.

Tab. 11.11: Cultivation parameters of the fermentation process.

Parameter Unit

Temperature 37 °C
Reactor type NFL 19L
Agitator speed Controlled, range: 200–600 rpm
Aeration rate 2 vvm (16 L/min)
p O2 set point 30%
pH set point 6.5
Filling volume 7,500 mL
Inoculation 500 mL, depending on ODpreculture

Tab. 11.12: Feed parameters of the fermentation process.

Parameter Unit

Growth rate, μset 0.04 h

Start biomass, c x,0 5.00 g/L
Yield, Yx/s 12.59 (KH2PO4) g/g
Start concentration, c s,in 5 (KH2PO4) g/L
 11.7 Managing the actual cultivation process | 289

1.6
1.4
OD 355 [nm]

1.2
1.0
0.8
0.6
0.4
0.2
0.0
0.000 0.002 0.004 0.006 0.008

cPO4 [mol/L]

Fig. 11.8: Exemplary calibration curve for a photometric molybdate phosphate assay to determine
the phytase activity.

The samples have to be stored on ice to avoid further cell growth. By measuring the
optical density in a spectrophotometer (UV/VIS) the cell growth and respective growth
phases can be monitored quickly. For Bacillus amyloliquefaciens it is sufficient to mea-
sure the sample at 550 nm using culture medium or deionized water as a blank. If the
optical density of the samples exceeds the linear measuring range, the samples have
to be diluted. The measured values for optical density can be related to cell dry mass
concentrations by a calibration curve. Hereby, the measured cell dry mass and optical
density have to be correlated. The cell dry mass concentration can be determined
gravimetrically by taking an appropriate sample volume. Cultivation parameters like
glucose, phytase, or phosphate concentrations are estimated as well. The samples
are centrifuged and the remaining cell pellets or the supernatants are analyzed by
photometric assays. An exemplary calibration curve for a photometric assay is given
in Figure 11.8.
However, other analytical methods than chromatography can be used as well to
determine these parameters. High performance liquid chromatography (HPLC), ion
exchange chromatography (IEC), or gas chromatography (GC) are worth mentioning at
this point. Figure 11.9 and Figure 11.10 represent the exemplary fed-batch process with

12 1.0
batch fed-batch

0.8
CDM [g/L]

8
μ [1/h]

0.6
CDM
μ
0.4
4
0.2

0 0.0 Fig. 11.9: Cell dry mass (CDM) in g/l and growth
0 5 10 15 20 25 30 rate (μ) in 1/h of an exemplary two stage fed-
tF [h] batch process with Bacillus amyloliquefaciens.
290 | 11 The practice of fermentation – a step by step guide through the workflow

8 0.4

a(phytase) [mmol/L*min]
batch fed-batch
c(PO4)

c(PO4) [mmol/L]
a(phytase)
6 0.3

4 0.2

2 0.1
Fig. 11.10: Phosphate concentration in mmol/l
0 0.0 and phytase activity in mmol/(l*min) of an
0 5 10 15 20 25 30 example two stage fed-batch process with
tF [h] Bacillus amyloliquefaciens.

all determined parameters over the process time. Additionally, the plotted values are
listed in Table 11.13. During the exponential phase within the first 8 h, cell dry mass in-
creased from 0.2 to 6.0 g/l with a maximal growth rate of 0.7 1/h. After a process time of
8 h the fed-batch process was started, reducing the growth rate to 0.04 1/h in the pro-
cess interval of 10 h to 26 h. The maximal achieved cell dry mass was around 10 g/l.
The phosphate concentration decreased from 6 to 1 mmol/l during batch phase and
slightly increased to 1.5 mmol/l at the end of the fermentation. The phytase activity
increased linearly after the fed-batch process was initialized till the maximum activ-
ity of 0.346 mmol/(min · l) was reached after 11 h. During the time from 16 h and 26 h
phytase activity remained constant around at 0.2 mmol/(min · l) due to the reduced
cell growth.

Tab. 11.13: Offline measured biomass concentration and phytase activity of the examplery two stage
fed-batch process with Bacillus amyloliquefaciens.

Process time Cell dry mass Growth rate Phosphate content Phytase activity
[h] [g/l] [1/h] [mmol/l] [mmol/(min · l)]
0.1 0.12 n.a. 5.99 n.a.
2.0 0.18 0.213 5.66 n.a.
3.0 0.27 0.409 5.46 n.a.
4.0 0.55 0.697 4.46 0.007
5.0 1.12 0.718 4.31 n.a.
6.0 2.20 0.677 3.40 n.a.
7.0 4.42 0.704 1.32 n.a.
7.8 5.73 0.329 0.85 0.043
8.9 7.98 0.299 0.75 0.167
10.5 8.60 0.047 0.67 0.285
11.2 8.79 0.029 0.92 0.346
16.1 10.12 0.029 1.17 0.193
24.2 n.a. n.a. 1.47 n.a.
25.1 9.75 0.054 1.63 0.183
26.0 10.17 0.045 1.46 0.171
 11.7 Managing the actual cultivation process | 291

When the fermentation process is finished, the remaining samples have to be ana-
lyzed, the reactor system has to be demounted, and the vessel including vessel com-
ponents, storage bottles, and sensors have to be cleaned. It is recommended to de-
contaminate the culture broth by another autoclaving step. The empty reactor vessel
and other peripheral parts have to be rinsed with cleaning agents like water, ethanol,
H3 PO4 , NaOH, or HCl. For each fermentation device, a suitable cleaning agent has to
be selected and applied in a defined amount and for a fixed time. Nearly all biopro-
cess fermentation devices require cleaning in place (CIP) operations. The necessary
time for CIP should be taken into account for scheduling a cultivation. A typical CIP
sequence might consist of the following steps: washing with water, rinsing with an
acidic solution, and washing with purified water.
In addition to fermentation results given in Table 11.13, an example protocol for
sample acquisition is given in Table 11.14. It can be used as a template to create a pro-
tocol for the measured values of taken samples during a cultivation.

Tab. 11.14: Example protocol for sample acquisition.

Sample Date Time Process Dilution OD CDM PO4 Phytase activity

no. time 550
JJ:MM:DD HH:MM [h] [−] [−] [g/L] [mol/L] [mmol/(min · l)]
0
1
2
3
4
5
6
7
8
9
10

In this chapter, it was shown how to manage an actual cultivation process in a lab scale
fermentation unit for the production of phytase from inoculation to data evaluation. It
has to be mentioned that this is an exemplary fermentation; depending on the aim of
fermentation, it is probably necessary to adapt the cultivation parameters and/or feed
parameters. However, by following the general steps in this section and the previous
sections it is possible to set up a fermentation in a bioreactor from scratch. We want
to encourage use of this book chapter as a practical guide to performing a cultivation
process, step by step.
292 | 11 The practice of fermentation – a step by step guide through the workflow

11.8 Integration on the process level – how to include the

fermentation into a whole process chain

Closing the fermentation chapter it should be reconsidered how the fermentation

stage interacts with upstream and downstream steps. Many examples have already
been mentioned in the previous chapters. There can be a direct physical integration of
process steps between upstream, bioreaction, and downstream. Even if this is not the
case, the stages influence each other in all directions, which the bioengineer has to
be aware of during the process development process. In best case, these interactions
can be anticipated to be of decisive advantage. Table 11.15 tries to give some order and
previously mentioned examples. Then some points are outlined in a bit more detail.
Strain robustness is an important issue. Some strains are in use only based on
historic development. Even simple things like shear stress resistance, stickiness, or
lack of byproducts could be targets. While in the lab complex media and supply of
vitamins are ok, for production cheap media are preferred. That has to be envisaged
already during strain selection and development. Additional pathways for certain sug-
ars is one way of making strains fit for the rough production environment or to keep
extracellular products in an inactive stable state. In other cases (e.g., microalgae) ex-
tremophiles can help suppress contamination. Many microorganisms with interesting
products have been screened and isolated, but they are hard to cultivate. Sometimes
they live in symbiosis with other microorganisms or work together in a conversion
chain. Cocultivation is an option currently discussed. Recall what the persistent prob-
lems using E. coli or Bacillus as expression systems are (auxotrophy, unnecessarily
complex metabolism, undesired proteases, foaming). Are artificial or minimal cells
delivered by systems biology a solution? The idea of robustness also drives attempts
to replace animal cells with phototrophic plant cells (moss, microalgae), which are

Tab. 11.15: Aspects of integration on the process level and on the level of process design.

Stages Aspect Measures and examples

Microorganism and cultivation Medium Media for lab and production
foaming
Control of physiology Process feeding strategy
Robustness
No byproducts
Microorganism and downstream Extracellular products Targeting tags
Cell wall permeability
Cell size and shape
Aggregation Auto/induced flocculation
Cultivation and downstream Cell recycle or retention High biomass concentration
Wastewater
Extracellular products
 11.9 Questions and suggestions | 293

more robust and can grow in pure mineral media. Intrinsic safety is also increased
because no viruses are known that infect plants and humans. This is of course a long
process triggered by recent success in genetic engineering towards humanized protein
products (glycosylation).
Extracellular products are a good idea to avoid intracellular inclusion body for-
mation and to ease separation. Citric acid or penicillin are established examples as
well as recombinant proteins, which may be naturally excreted or engineered with a
targeting tag. In the case of citric acid cell wall mutants are employed. Cell disruption
or extraction also turn out be problematic in some cases. Having a look at these unit
operations with respect e.g., to the cell wall can be done already during strain devel-
opment. In situ extraction during the cultivation with an immiscible extractant has
been tried out.
In the baker’s yeast example fed-batch processes in the cultivation stage are nec-
essary to cope with the biological features of the Crabtree and Pasteur effects. Fur-
thermore, the process strategy could influence product quality and, in the context of
integration on the process level, filtration efficiency. As an aside, budding of yeast
also influences the brewing process, be it bottom fermented (sedimentation) or top
fermented (flotation). Back to baker’s yeast, what else could be done during strain de-
velopment e.g., by genetic engineering (if it was allowed) to make the process simpler
and more intensive? Yeast strains are the result of 100 years of strain development.
This virtual glass bead game could include overexpression of the respiratory chain.
Yeast can do that in principle as we have seen during aerobic growth on ethanol. Some
strains cannot produce ethanol. Is it sensible to use such strains, when some yeasts for
single cell protein (SCP) would come into question? This is obviously not expedient,
as we need the fermentative pathway to let the dough rise. So an inducible alcohol de-
hydrogenase (ADH) would be an option for switching on under the specific conditions
in the dough. Similar considerations for other expressions systems are made.

11.9 Questions and suggestions

1. Antifoam agents often have a negative effect on the cultivation. What it is the
reason for this problem and what are some possible solutions?
2. The electrical engine has 1.1 kW power and max 1500 min−1 stirrer speed.
Check it against the needs from Chapter 5 on bioreactors.
3. Draw a reactor scheme by heart and indicate the parts of the reactor that have
to be steamed and which valves have to be opened/closed in which sequence.
4. The oxygen transfer rate (OTR) is often the limiting step during a cultivation.
How can it be increased in a running process?
5. In which way is the k L a value increased in the present cultivation and why is
the parameter limited to a specific range?
294 | 11 The practice of fermentation – a step by step guide through the workflow

1. By adding (silicon oil based) antifoam agents to the culture medium several prob-
lems can occur. The mass transfer rate can be decreased due to a change in surface
tension. The agents may also be metabolized by some microorganisms. Especially
for enzyme production specific antifoam agents behaving inert to the microor-
ganism are preferred. To overcome these disadvantages mechanical foam break-
ers like e.g., centrifuges, cyclones, and impellers can be used. Unlike chemical
agents mechanical foam breakers destroy foam after it has been formed. Another
alternative is the use of physical methods like ultrasound, or electrical or thermal
treatment.
2. The OTR can be increased by:
– mass transfer coefficient (k L ): decreasing boundary layer by stirring faster
– PO2 : increasing oxygen partial pressure in in-gas
– interface area (a): increases through faster stirring → smaller bubbles
4. The k L a is controlled by the stirrer speed. The range is limited to a lower and upper
value. The lower value is dependent on homogenization of heat and mass. The
upper value is capped to avoid damaging of the cells due to shear stress.
12 Modeling – art and handcraft of mathematically
describing bioprocesses
Modeling is not an esoteric task we can do at the end of a process development as a
cherry on the cake. In fact, it is an integral part of process design and has to be done
iteratively from step to step during the design process. The willingness to understand
inner relationships of systems and to effectively design processes brings us inevitably
to modeling. In this chapter some general ideas about modeling of bioprocesses are
collected. Specific features to model metabolic networks are one kernel of a biopro-
cess model. Modeling itself is a structured and integrated processes following specific
ideas, structures, and rules for the different aspects we meet in biotechnology. As gen-
erally applicable examples aerobic, anaerobic, and phototrophic processes are given
in some detail. The examples also give hints to different modeling techniques setting
the follower in a position to build models for other or more complex bioprocesses.
Last but not least, quantitative analysis of the model and model outcomes have to be
computed, a process called simulation. No specific mathematics is needed; a short
introduction will enable us to work with the models in the supplementary material.

12.1 Modeling – what it might be and what it is good for

All engineers are working with models sometimes without being fully aware of it. The
imagination of electrical or gravitation fields are model representations allowing us to
quantitatively design technical devices. This concept has been applied very success-
fully even knowing that reality may be more complex. Behind all scientific terms stand
structured ideas, making up their impact on scientific thinking. Before being lost in
details we can try to formulate an admittedly quite broad definition:

Models are descriptions of natural and artificial systems in a formal representation space.

Modeling is hence an integrative approach between different sciences or people who

contribute their specific knowledge. More attributes filling this formal statement with
life are made clearer in the next paragraphs. In our daily lives we encounter models,
often graphical models. Street maps can be taken as a common example. Comparing
such a map with a satellite picture showing the same section of a landscape, the differ-
ences become clear. First of all, many bewildering details are left out, while the focus
is on showing the streets only. So simplification or better said abstraction is one ma-
jor issue in setting up models. The different streets like highways or country roads are
kept in different colors. Of course in reality streets appear in gray. This difference be-
tween graphical model and reality is a tribute to applicability, the second basic issue.

https://doi.org/10.1515/9783110315394-012
296 | 12 Modeling – art and handcraft of mathematically describing bioprocesses

Orientation on usefulness defined by model builder or model applicant is a pragmatic

consequence. In biology, graphical models are also a classic method. A drawing of a
cell with several details is based on many microscopic pictures, none of them alone
showing all the details of the drawing. Analog models can be a useful means of com-
munication. An example in bioprocess engineering is the bottleneck model. Behind
every model purpose stands people who will work with the model. This has to be con-
sidered by the model builder. Liebig’s barrel model is a good example of user oriented
graphical modeling.
In engineering graphical representations are also a common method, as process
flow sheets prove. To go more into details and to describe invisible relations between
parts of a system mathematics is the language of choice. Consequently, in the context
of this book we are interested in mathematical models, especially process models. This
leads us to a third issue: complexity. The decision which and how much information
is lumped into the model has its limits on the one hand in known details and on the
other hand in the purpose of the model. Figure 12.1 gives an overview of models of
different complexity based on data and knowledge.
In control theory information is needed to adjust parameters of the controller,
e.g., PID control. Often there is not enough knowledge to directly calculate dynamic
responses e.g., to disturbances. In such cases simple step response functions are mea-
sured and used to set up basic system representations as P, PI, or PID systems. Such
simple models are called data driven models. Despite the fact that no mechanistic
assumptions about the system are made, such models are good enough to fulfill the
valuable task of control. Neural networks are another more complex example. In mul-

New data
A priori assumpons Data New hypotheses

Model Improved
experiments

Controller Simulaon Deeper

design scale-up knowledge
opmisaon

Increasing Structured
Data-driven models informaon content mechanisc models

Fig. 12.1: Flow sheet to show generation of structured models from simple ones and increasing infor-
mation content by cyclic running through the modeling process.
 12.1 Modeling – what it might be and what it is good for | 297

tifactorial experimental design a quadratic cost function is the simple model back-
ground on which the experiments are planned. The inclusion of more information
leads to moderately structured models. The simple growth model we used until now
is an example of this. It includes some assumptions about uptake kinetics and growth
yield but without overstressing any details. Putting more and more knowledge into the
model will presumably lead to more precision and more reliability of such a mecha-
nistic model. The highest meaning of a model, although it is assembled from already
known bricks, is to help get new insight into a system and to gain better understanding
of the system as a whole.
Models in the context of bioprocess engineering are meant to be employed for pro-
cess development. This includes plant simulation, optimization, media design, and
scale up. Changing process policy e.g., from batch to fed-batch needs precise under-
standing of the interaction between the medium and the cell. Exact kinetics for most
of the medium compounds are necessary. Simple heuristic models, valid only for a
specific cases, are not adequate. The intracellular structure also has to be considered
in order to understand constraints and flexibility of the metabolism, because intra-
cellular bottlenecks and stoichiometry will dominate the whole process. Models for
successful process development are therefore structured models. Validity as the next
important issue increases with inclusion of mechanistic items. In the best case a model
is equipped with clearly defined limitations of validity.
The framework to start (Figure 12.2) is definitely conservation laws as they is stan-
dard in chemical engineering in general. This holds especially for the reactor equa-
tions. Known stoichiometry, and macromolecular and elemental composition will be
added in a next step on the cell level. The model building process is a structured pro-

Inial state Intermediate stages Final state

Model evoluon

Heurisc Physiological Determinisc

components components components

Fig. 12.2: Framework for building up a model from different model building blocks with different
deterministic, mechanistic, or statistical background.
298 | 12 Modeling – art and handcraft of mathematically describing bioprocesses

Reactor level Phase level Cell level

S

Gas-
phase
rS
yX,S
Liquid-
X
phase rX = μ

Bio-
rP
phase

Fig. 12.3: Breaking down of a reactor model to submodels, here the physical phases; a metabolic
submodel of the biophase is further indicated.

cedure with formal metarules on different levels and model aspects. Facing the com-
plexity of a cell and the whole process, it is clear that model components will always
be missed. Depending on the purpose of the model we can be satisfied or look for fur-
ther relevant information. Even here the model as developed so far can be useful. It
supports model based experimental design to bring out as much information as pos-
sible to allow for a quantitative description of missing parts. Finally the model can be
regarded as a complete description of the process.
What happened until now in this book? Process models have already been set up
more or less intuitively. But most important is also that a structure has been assigned
to the problem of model setup. The system was composed of three subsystems, namely
gas phase (as described in Chapter 5 in some detail), the liquid phase, and the ‘bio-
phase’ (Figure 12.3). All three phases of course have mutual mass transfer and other
ways of exchange. In the liquid phase usually no or well known reactions take place
allowing us to apply deterministic rules of mass and energy balances.
These equations (12.1a) we called the reactor model:

dc
= ±q × c ± r(c) (12.1a)
dt
This well known reactor model (here given in bold as vector representation) consid-
ering accumulation, transport, and reaction has to be amended with a model for the
biophase. These were based basically on observations. For substrate uptake a satu-
ration curve is observed, which we interpreted, maybe a bit prematurely, as enzyme
kinetics. Reducing several substrate uptake systems of many microorganisms to only
one is here a simplification. Further data show that growth is linearly dependent on
substrate uptake, which was interpreted as metabolic energy balance. In compact vec-
tor form (Equ 12.1b) this reads:

rkin = f (c) ; r = y × rkin ; (12.1b)

 12.2 Predator-prey model | 299

Information processed in the physiological model is not really deterministic but more
heuristic and the related parameters taken as statistical values from measurements.
There remains the concern that the way we came to the model was a bit arbitrary and
would fail in more detailed and complex biological systems. Modeling should be a well
structured approach, which we want to go into now. We start with a classic example
from ecology.

12.2 Predator-prey model

In the beginning of the last century the Hudson’s Bay Company observed that the num-
ber of sold furs of snowshoe hares and lynxes was affected by strong fluctuations of
several years duration. Concerned about the severe impact on economics they tried
to understand the phenomenon following the work of the chemist Alfred Lotka. He
worked on oscillation in chemistry, society, and ‘physical biology’. To follow his ideas
we first have a look at the data (see Figure 12.4) and find the reasons for the fluctua-
tions. Observations are that the frequency of the oscillations is more or less constant
as well as the mean values. Further, the peak of the number of the furs of the hare sets
in earlier than the one of the lynx. These observations are not enough for understand-
ing but may act as a benchmark after we uncover the inner relations between lynx and
hare: basically the relation of predator and prey.

180
Lynx fur/1000

160

140

120

100

80
Hare fur /1000

60

40

20

0
1850 1860 1870 1880 1890 1900 1910 1920 1930
Year

Fig. 12.4: Original dataset from Hudson’s Bay Company showing the number of sold lynx and hare
furs.
300 | 12 Modeling – art and handcraft of mathematically describing bioprocesses

The first steps are to define the system boundaries and to list the most important vari-
ables involved. In our example it is the scope of Canadian northern forests and the
number of lynxes NL (t) and hares NH (t). The dataset represents an inherently discrete
time process with discrete values. Nevertheless, we know that the number of animals
is very high and that they propagate also during the year, so we are encouraged to
use a time continuous model. Remember that a similar transition has been done al-
ready describing microbial growth. Here we keep the letter ‘N’ understanding that is
a real value with the meaning of an areal population density. In the next step the in-
terrelations between each of the variables with themselves as well as with the other
variables have to be fixed. Therefore, we formulate some model hypothesis and find
mathematical representations for it:
The undisturbed growth of the hare population follows the same idea as the
growth of microorganisms. The ‘undisturbed’ growth of the lynxes is characterized by
slowly dying in case they have no feed. We define the decay rate positively, but will
consider the decay with a minus sign in the final balance:

dNH 󵄨󵄨󵄨󵄨
󵄨 = RH,growth = rH,growth ⋅ NH ; (12.2a)
dt 󵄨󵄨󵄨undis
dNL 󵄨󵄨󵄨󵄨
󵄨 = RL,death = rL,death ⋅ NL (12.2b)
dt 󵄨󵄨󵄨undis

The structure of this simplest possible linear system, which includes a feedback from
the state variable to its derivative, is common in nature and technology at least in
simplifications. Other examples are radioactive decay, chemical reactors with a posi-
tive temperature coefficient (can even explode!), or the Lambert–Beer law (flux of ab-
sorbed photons per path length is proportional to photon flux).
Something happens only if lynx meet hare, which is the first interplay. The prob-
ability is higher the higher the population densities of the respective species. This can
be fixed by introducing a probability coefficient wH,L . With a given ‘efficiency’ in hunt-
ing by the lynx, in the model represented by a yield coefficient y−H,L , this ends up with
the death of the hare. The other way around, the lynxes need a given amount of caught
hare for reproduction, noted as yL,H .

dNH 󵄨󵄨󵄨󵄨
󵄨 = RH,meet = y−H,L ⋅ wH,L ⋅ NH ⋅ NL ; (12.3a)
dt 󵄨󵄨󵄨meet
dNL 󵄨󵄨󵄨󵄨
󵄨 = RL,meet = yL,H ⋅ wH,L ⋅ NH ⋅ NL (12.3b)
dt 󵄨󵄨󵄨meet

Take notice of the analogy to the mass action law, where the same probability ar-
gument leads to a comparable formulation. The parameters y and w appear only as
a product and are therefore linearly dependent. Further, they cannot be determined
from other observations, so we lump them in this model to the new parameter k, sim-
ilar to the reaction constants in chemical reaction technology. Remember also this
lumping process in the derivation of the k L a value. Now we can set up a complete
 12.2 Predator-prey model | 301

200
Lynx fur/1000

180

160

140

120
Hare fur/1000

100

80

60

40

20

1840 1860 1880 1900 1920 1940

Years

Fig. 12.5: Simulation of the predator-prey model with the parameters: rH,growth = 0.5; rL,death = 0.4;
w H,L = 0.01; y −H,L = 1.0; y L,H = 0.6; N H,0 = 20; N L,0 = 40 (given in thousands).

population balance for both species as sum over all of their single rates ΣR :

dNH
= RH = rH ⋅ N H − k H ⋅ N H ⋅ N L ; (12.4a)
dt
dNL
= RL = k L ⋅ N H ⋅ N L − rL ⋅ N L (12.4b)
dt
A simulation example is given in Figure 12.5. Indeed, the periodicity and the other
features of the measurements are fairly well represented.
A simulation program is given in the supplementary material. Such population
models are also called Lotka–Volterra models, according to the already mentioned
Lotka and the Italian mathematician Vito Volterra, who later described similar ob-
servations for fish populations in the Mediterranean Sea. The impact of such models
goes beyond population dynamics. Activator-inhibitor systems follow the same pat-
tern. Such systems are for example biceps and triceps working against each other
to allow a stable movement of the arm, a principle called antagonism. Pairs of hor-
mones often work according to this principle like insulin and glucagon or adrenalin
and noradrenalin. Such activator-inhibitor systems may be stable or unstable. A con-
tinuous cultivation can also be understood in this way. Substrate is the activator while
‘undisturbed’ feed leads to substrate accumulation in the reactor. This supports in-
creased substrate uptake and growth and finally more biomass. Biomass as inhibitor
leads to negative feedback on the substrate concentration. We have already seen that
for substrate inhibition the sign of feedback changes to positive leading potentially to
an unstable situation. Other examples of cyclicality are business or climate cycles. In
302 | 12 Modeling – art and handcraft of mathematically describing bioprocesses

most cases oscillations are not so desirable other elements like damping factors. More
complex predator-prey systems tend to be more stable. Lynxes may have other prey
than hares and vice versa, which turns out to be a stabilizing factor. That makes it un-
derstandable that oscillations are observed in nature mainly in species poor regions.
Mathematically speaking stable cycles require a nonlinear element, like the product
of the two species, and a positive feedback loop. The undisturbed growth equation is
such a loop as more hares support development of more hares exactly as in microbial
growth.
Did we really find a basic description of reality in population dynamics and is our
model hypothesis proven? Doubts are definitely appropriate. First of all the number of
furs is in the best case proportional to the real population density and, if not, with the
same factor for both species. It is indeed unbelievable that on average only three times
more hares exists as lynxes and that the predator can live by catching only three prey
animals per year. Furthermore, human hunters may have an influence in reducing the
number of both species with different efficiency. Could an external influence control
the systems like solar cycles (sunspots)? Some people found the time to exclude this
specific idea. As many experts think now, the real predator-prey system is the one
between grass and hares being now the predator. The number of lynxes follows the
cycle passively; so much for model validity.

12.3 Linear networks – finding stationary flux equations

After we have identified activator-inhibitor systems as structures of possible interest to

describe biological systems we seek other general patterns. Many systems have a net-
work structure. We know them as electrical or computer networks. This may be traffic
systems like railways or streets. Also rivers, water supply networks, or leaf veins fall
into this category. Here the question of interest are fluxes (transport) along edges like
number of cars per second driving along a street or the water flux in a river. These
fluxes interact with each other and the points where that happens are called nodes
of the network. A node can be the mouth of a small river into a bigger one. The in-
terconnection would be simply an addition of both water flows upstream to the one
downstream. A more complicated view could be the temperature or pollution load of
the rivers. Typically these nodes are connected only by a few edges (arcs, pathways)
with other nodes thus forming meshes. This structure makes final modeling easier
than dealing with a lot of variables, all of which are interconnected. Another simpli-
fication is to ignore all kinds of storage. A parking place at a highway will not change
much the number of cars per time unit between a motorway access and the next exit.
Metabolic networks are of highest interest in the context of bioprocess engineer-
ing. Beyond a graphical metabolic structure showing metabolic pathways and their
mutual links, the metabolic fluxes along the pathways should be calculated and made
visible. We did this already on a strongly simplified basis by calculation of the specific
 12.3 Linear networks – finding stationary flux equations | 303

rCO2
Nresp NADH2
CO2
NAD
rCat
rS H2O
Glucose rH2O
Pyruvat ADP+Pi
ATP
NCat rAna
NAna
rNH3
rX

NH3
Biomass

Fig. 12.6: Metabolic structure of a typical aerobic microorganism; the circles are possible subsystem
boundaries. N stands for node.

turnover rates. Accumulation of metabolites may have biological meaning but are ig-
nored for now. This stationary view leads to sets of algebraic equations. The simple
approach, called ‘physiological model’, consisting of Michaelis–Menten kinetics for
substrate uptake and of a Pirt equation was a first attempt based on observation. Now
it is time to find a stringent and unified procedure for setting up such models for more
complicated cases. As a first example the aerobic metabolism as depicted in Figure 12.6
is chosen.
For complex metabolic networks we have to set up a structured approach for find-
ing suitable balance boundaries and setting up the balance equations from network
analysis. Here are important rules:
– Each node to be considered has to be located at least inside one system boundary.
Otherwise the stoichiometry of this node is not represented in the model. In the ex-
ample three nodes are defined for the catabolism Kcat , the respiratory chain Kresp ,
and the anabolism KAna . Consequently, three system boundaries can be identi-
fied, here shown as ellipses (brown).
– The number of system boundaries equals the number of metabolic nodes. The def-
inition of an additional system e.g., around the whole cell may be useful for later
model validation, but does not give additional information besides the balances
around the individual nodes.
– Each metabolic path has to be intersected (at least) by one balance border. Oth-
erwise it will not appear in the respective balance equations and therefore not in
the model.
– The maximum number of linear independent balance equations nmax,knot for one
node cannot reach or exceed the number of nodes nknot . Otherwise the node would
determine itself completely and would be decoupled from the remaining part of
the network and the outer world.

These rules are also applied to networks with several hundred nodes. Besides, weaker
linear relations are also observed in living cells. These are not directly stoichiomet-
304 | 12 Modeling – art and handcraft of mathematically describing bioprocesses

rically determined but have their origin in intracellular energy or material balances,
which are subject to change by the cells according to the environmental conditions.
Metabolic fluxes interact in specific ways corresponding to specific approaches to find-
ing model equations. This will be exercised now for the aerobic metabolism.

12.4 Aerobic growth – setting up a first general model

The first step is to look at the metabolism and to select important metabolic pathways.
For this example the metabolic structure in Figure 12.6 is chosen. Fluxes are counted
positively in the balance if they go inside a node, and negatively going outside. That
does not mean that they go in reality (always) in this direction. The numerical value
of a flux will then have a negative sign.
Substrate (glucose) as C6 body is taken up and converted via glycolysis to pyru-
vate (C3), which stands here as representative for all metabolites of the TCC. The two
ATP produced are neglected here. The same holds for NADH2 , as it compensates more
or less with redox demand in growth the remaining parts are small compared to re-
dox production in respiration. So the virtual metabolite ‘Cat’ has the same degree of
reduction as glucose. The brown circles represent system boundaries for setting up
balance equations. All parts having to do with respiration, including production of re-
dox equivalents and CO2 as well as oxygen uptake and ATP generation, are lumped in
one subsystem. The anabolic pathways include biomass formation from metabolites
in the TCC and nitrogen uptake as a model also for other nutrients.
A first approach is setting up material balances and stoichiometric relations. El-
emental balances are a specific form of mass balances. Balance equations can be set
up either by looking at mass or at molar fluxes. The latter option has the advantage
that some side reactions like CO2 or H2 O addition can be ignored for a while. On the
other hand, a conversion factor from the molar growth rate to real biomass has to be
found. We start here with mass flux balances as they link to the representation used
in the other chapters. For the mass flux balance around the TCC all three fluxes have
to be summed up. All fluxes are given in a condensed representation. The first one is
the flux through glycolysis. In the TCC precursors for growth are produced (rAna ) and
redox equivalents (rCat ) for respiration. Outgoing fluxes rAna and rCat for growth and
respiration are counted negatively, leading to rS − rCat − rAna = 0; pyruvate does not
accumulate. This is actually the first row in Equation (12.5). Note that no additional
knowledge about yields is required. A second equation could be allowed, but would
result in redundant information. Each elemental balance for example would lead to
a linearly dependent equation. From a biochemical and physiological view it is clear
that the organisms can allot the glucose arbitrarily into respiration as energy source
or into growth as carbon source according to the actual needs. Anaplerotic reactions
in the TCC help them to do so.
 12.4 Aerobic growth – setting up a first general model | 305

mass balance NCat

[ mass balance N ]
[ Ana ]
[ ]
[ N balance NAna ]
[ ]
[ C balance N ]
[ Resp ]
[ ]
[ O2 stoich NResp ]
[ ]
[H2 O, CO2 stoich NResp ]
[ ATP balance ]
1 0 −1 −1 0 0 0 0 rS (12.5)
[ ] [ ] 0
[0 0 0 1 1 −1 0 0 ] [ rO2 ] [ ]
[ ] [ ] [0]
[ ] [ ]
[0 0 0 0 MN
−eN,X 0 0 ] [ rCat ] [ ]
[ MNH3 ] [r ] [ [
0]
]
[ −MC ] [ Ana ]
] = [0]
MC
= [0 0 0 0 0 0 MCO2 ] ×[
[ MCat ] [rNH3 ] [ ]
[0 −3
0 ] [ ] [0]
] [ rX ] [ ]
1
[ 0 0 0 0
[ MO2 MCat
] [ ] [ [0]
]
[0 0 0 0 0 0 1 −1 ] [ ]
[ MH2O MCO2 ] [ rH2O ]
0
[0 0 16
MCat 0 0 −yATP,X 0 0 ] [ rCO2 ] [ ]

In the same manner the mass flux balance for the metabolism is set up (second row),
meaning that the anabolic mass flow plus the nitrogen source (representative for other
inorganic salts) delivers the final biomass flux, which is actually the specific growth
rate. Assuming constant biomass composition, nitrogen uptake is regarded as propor-
tional to the anabolic flux; compare the medium calculation in Chapter 3, leading to
the third row. Now we have two equations for the metabolic node with three unknown
fluxes, which is sufficient and the maximum according to the rules. Respiration is a
completely defined process without any degree of freedom. So three additional equa-
tions (row 4 to 6) for four unknowns based on oxidation stoichiometry can be set up.
The carbon balance is set up first, based on the fraction of C in Cat and CO2 . Stoi-
chiometry can be best formulated on a molar basis. So fluxes are divided by their mo-
lar mass (M). The virtual metabolite Cat (C3 body) needs three O2 to be completely
oxidized. Row 6 says that per mole CO2 one mole H2 O is produced. Now six equation
for eight unknown fluxes have been set up.
Basically, the allotment of glucose to respiration or growth is one degree of free-
dom. As a biological hypothesis we assume that the cell will use all ATP generated in
respiration for growth. On the one hand, growth requires a thermodynamically given
amount of energy, while on the other hand, ATP from respiration will not be wasted.
So setting up the ATP balance sounds sensible. Complete oxidation of glucose is fairly
well known to deliver 32 ATP/mole, we therefore set 16 mole ATP per mole Cat. ATP
demand for growth can only be roughly estimated from references, so we use a yield
coefficient yATP,X meaning the amount of ATP in moles to allow for building up 1 g of
Biomass. Now we have seven equations: the maximum for the whole cell. Only one
possible equation is left to describe the interaction of the cell with the environment in
the form of kinetics. In the example it could be either substrate, oxygen, or nitrogen. In
the following we look at substrate uptake kinetics as the limiting step. All other fluxes
306 | 12 Modeling – art and handcraft of mathematically describing bioprocesses

follow, then the glucose flux. If the nitrogen source or oxygen is limiting, the cell has
to downregulate the substrate uptake respectively. The metabolism of the aerobic cell
is indeed highly stoichiometrically determined.
The matrix describing the intracellular linear dependencies is called the system
matrix or more specifically for our biotechnological purpose the stoichiometry ma-
trix Y. The dimensions of the matrix are nequ rows (number of equations) and nrates
columns (number of fluxes). Of course nequ < nrates . In our example for aerobic growth
nrates = 8 and nequ = 7. The difference (here 1) is the number of degrees of freedom;
the cells have to react to the environment. This is measured with the rank(Y) of the
matrix. Calculation (see virtual supplement) in the example case delivers 7, so one de-
gree is left, meaning one additional equation has to be and can be set up. It can also
turn out that in complex networks the matrix is not of full rank. In that case the rank is
lower than the number of equations. That might happen in cases where the metabolic
network under investigation has equilibrium reactions or anaplerotic sequences or cir-
cumvents the rigid stoichiometry. During setting up the model equations these cases
may not be detectable intuitively. So looking at the rank of the stoichiometry matrix
delivers a valuable outcome to understand better the physiological behavior of the
cell. To sum up, a generic stoichiometric model can be written as:

Y × rT = 0 (12.6)

This contains the whole information with respect to the intracellular stoichiometry.
The degrees of freedom nfree = nrates – rank(Y) can be filled up with kinetic equations
conveying between intracellular and extracellular fluxes.

1.0
Normalized flux values

0.8

0.6

0.4

0.2

0.0
rS rO2 rCat rAna rNH3 rX rH2O rCO2

Specific rates

Fig. 12.7: Flux distribution in an aerobic cell; the calculation is based on the stoichiometry matrix
given in Equation (12.4a).
 12.5 Anaerobic growth – traps and pitfalls | 307

Finally the model is used to find a value for the a priori unknown ATP demand of
growth. From numerous observations we know that approximately only half of the glu-
cose is used for growth and the other half is respired, which leads to y X,ATP = 18/90 ≈
1/yCat,ATP . In Figure 12.7 the flux distribution for our example is shown as a bar graph.
The values are normalized for rS = 1. This is possible as the relations do not change
with increasing or decreasing substrate uptake.

12.5 Anaerobic growth – traps and pitfalls

In anaerobic pathways things look a bit different but more simple on the first glance
as shown in Figure 12.8. Substrate is degraded via glycolysis delivering two ATP per
glucose. Note that two ATP per glucose are needed firstly to form phosphorylated com-
pounds, while each one of the two resulting C3 bodies produce two ATP per mole. Fur-
thermore, two redox equivalents per mole of glucose are generated. Now we feel en-
couraged to simplify the situation by neglecting possible byproduct formation. As the
amount of available ATP is much lower than in the aerobic case, only a small amount
of metabolite ‘Cat’ can be used for growth. The remaining part cannot simply be ex-
creted because NADH2 is produced in excess. Only a small amount according to the
low growth rate is needed. The predominant part has to be transferred to ‘Cat’ again,
and finally ethanol is excreted.
Now the stoichiometric matrix (12.7) can be set up from mass balance around the
central metabolism followed by the energy and redox balance:

mass balance NCat 1 −1 −1 rCat 0

[ ] [ ] [ ] [ ]
[Redox balance NCat ] = [ 1 − M1X 1 ] ]
− MFerm ] × [ r X ] = [0]
[ [ (12.7)
[ ] [ MCat ]
1 −18
[ ATP balance N Cat ] [ MCat MX 0 ] [rFerm ] [0]

NAD NADH2
CO2
rCO2
ADP ATP

rS Ncat
rEth
Glucose rCat Pyruvate Ethanol
rFerm

rAca
rGly
Acetaldehyd
Glycerin
rX

Biomass

Fig. 12.8: Simplified structure of an anaerobic metabolism. The dotted fluxes are switched on or off
by the cell depending on present needs.
308 | 12 Modeling – art and handcraft of mathematically describing bioprocesses

This results in three equations for three unknown fluxes. That is forbidden as overde-
termination, as no degree of freedom is left to react to the environment. But there is
a way out of this dilemma looking at the rank of the matrix. Rank (Yanaer ) can be 2
when the redox demand of the cell exactly matches the produced amount from the
metabolites allotted to growth. In the matrix it can be seen that the first and the sec-
ond row are linearly dependent. The two rows can be transformed into each other by
a constant factor for the case that MCat = M X and MFerm . During model simplifica-
tion that was actually supposed as C3 bodies with the same degree of reduction. That
means in particular that the organism has to have the same degree of reduction as
the substrate glucose. That is of course physiologically speaking not sensible. Lipid
accumulation would for example not be possible. In any case the cell needs further
degrees of freedom. For anaerobically growing yeast and anaerobic bacteria nature
has given us this degree of freedom via acetaldehyde and glycerol excretion. Per mole
glycerol one NADH2 can be ‘disposed’. As a phosphorylated compound (dihydroxy-
acetone-phosphate) is used for this purpose, the process is at the cost of one ATP. The
usually high maintenance leads to a change in ATP/NADH2 ratio, making balancing
via byproducts necessary. Furthermore, acetaldehyde can freely diffuse through the
cell membrane leaving NADH2 behind. Diffusion depends on the concentration gradi-
ent over the cell membrane, making acetaldehyde flux not only dependent on growth
rate but also on biomass concentration and process feeding strategy with feedback
on glycerol excretion. Byproduct pattern is therefore difficult to understand simply by
looking at the data. What we learned from this model is that byproduct formation in
anaerobic processes is not only necessary to give beer a good taste, but is a biological
necessity. In more complex situations the only way to get insight into the behavior of
cells is using models.
The examples we investigated until now give rise to the impression that cells are
highly chained by all kind of compulsions as indicated in Figure 12.9. Thermodynam-
ics, here in the appearance of NADH2 and ATP balance, or mass conservation in the
form of elemental balances, form a tight corset. That allowed us as practical spinoff
to assign flux values also to intracellular fluxes. Based on modeling it is possible to
look one step deeper into the metabolism than is possible only based on extracellular
measurable metabolic fluxes.
But life is much more flexible. We already looked at examples of anaplerotic se-
quences and switching on/off additional pathways. In other cases cells have two par-
allel pathways for different purposes, but possibly compensating each other. Geneti-
cists are sometimes astonished by how good some microorganism can compensate
knockouts. Glucose can be degraded via glycolysis or via the pentose phosphate path-
way. As there is no rigid stoichiometry a prediction about the fluxes through these
two pathways is not possible. An approach from metabolic analysis to measure such
fluxes directly is using C13 labeled substrate compounds. The principle is shown in
Figure 12.10.
 12.5 Anaerobic growth – traps and pitfalls | 309

O2 X
Product
Element Reductants
sources energy
APT balance

Unknown balance
Liming
steps
Stoichiometry s
The law
rmo
dynam rvaon
ic fr a m e c o nse

Fig. 12.9: Different material fluxes into and out of the cell are strictly constrained by thermodynamic
conservation laws and stoichiometries.

As a substrate example a C5 compound is shown at the top of the picture; the carbon
atoms are counted from 1 to 5. The two colored carbons (C1 and C3) are labeled. The
colors (no difference in reality) are used to follow their fate on the way through the
metabolism. In pathway 1 C1 and C2 are cleaved, while in pathway 2 the first three
carbons including the labeled ones are kept. Both remaining C3 bodies are in chemi-
cal equilibrium allowing no stoichiometric statements. Now it is possible to measure
the labeled carbons in the product using NMR technology. The result is that C1 of the
product (mixture of the former C1 and C3) is completely labeled while C3 (the former
C5 and C3) is only labeled to 50%. This allows the conclusion that half of the substrate
is degraded via pathway 1 and the other half via pathway 2. This information about un-
measurable intracellular fluxes is a powerful tool to be included in linear flux models
or to verify them.

C1 2 3 4 5
Measurable Substrate known
substrate flux C13 labeling
rmeas

Path 1 Path 2
C3 4 5 Intracellular C1 2 3
metabolites
Measurable
product flux Path 3 Intracellular
fluxes rin

Product

Fig. 12.10: Principle of flux measurement using labeled carbon sources.

310 | 12 Modeling – art and handcraft of mathematically describing bioprocesses

12.6 Back to baker’s yeast – more degrees of freedom

In Chapter 7 about fed-batch processes we got to know something about Pasteur and
Crabtree effects. Under oxygen limitation or above a certain growth rate, yeast is able
to produce ethanol parallel to respiration. According to the spirit of modeling it is
not enough to formulate these relationships as an “if . . . then . . . else” rule. Rather
we want to find a mechanistic reason either on the reactor (oxygen transport) or on
the metabolic (limited respiratory capacity) level. In order to approach the problem
from the outside, a dataset of a continuous cultivation (X/D-diagram) is given in Fig-
ure 12.11 (a) and (b).
For small dilution rates the curves for measured concentrations look as expected.
Above a critical growth rate D = r X > r X,crit ethanol formation sets on. It is understand-
able that ethanol production is at the cost of substrate and biomass, but otherwise
the curves look a bit unusual. On the right hand side the calculated specific turnover
rates are plotted. The onset of ethanol formation is also visible by the increasing res-
piratory quotient (RQ value) and coincides with the maximum of the specific oxygen
uptake rate. Considering this maximum could be an idea to follow up when setting up
a model. In fact, a bottleneck is a good working hypothesis to start with.
Yeast catabolism, shown in Figure 12.12 as a strongly simplified graphical model,
is a merger between the aerobic and the anaerobic case. From the viewpoint of mod-
eling we now have a problem. There are four metabolic fluxes but only two equations
can be formulated. One degree of freedom is necessarily used by the cell for the link
between growth and energy formation. This leaves one degree of freedom for the cell
to channel carbon into respiration or the fermentation pathway.
What makes the yeast cell produce ethanol with low energy gain instead of only
respiration? We cannot follow all intracellular control mechanisms on the genetic or
epigenetic level, which would be the pure mechanistic view. A successful approach
to cope with this situation is to anticipate the control goals of the cells and model
the resulting effect. In continuous cultivations it is not necessary to consider control
dynamics. In several published metabolic flux models the control goal has been stated
quite generally as: “The cell utilizes all degrees of freedom to maximize growth.” The
formal notation (12.8) reads:
󵄨󵄨
󵄨 cS
max {r X 󵄨󵄨󵄨Y × r = 0, rS ≤ rS,max ⋅ , r X ≤ r X,max } (12.8)
󵄨󵄨 kS + cS

The problem description is interpreted in the following way: The cells try to optimize
their individual specific growth rate r X . This is first of all a linear optimization prob-
lem. Without a specific cost for increasing μ, which is the case here, it would go to
infinity. But the cell is subjected to the linear constraints given by the stoichiometric
matrix and the possible substrate uptake rate. The less than or equal sign is used here
to allow the cell to downregulate substrate uptake in case it brings an advantage, e.g.,
to prevent a glucose overflow. Additionally, we are not sure that substrate uptake can
 12.6 Back to baker’s yeast – more degrees of freedom | 311

40 12 6
14

CPR [g⋅L-1⋅h-1]
10
12
30
8 10 4
Substrate [g⋅L-1]

Biomass [g⋅L ]
-1
Ehanol [g⋅L-1]

8
20 6

OUR [g⋅L-1⋅h-1]
6
4 2
10 4

2
2

0 0 0 0
0.0 0.1 0.2 0.3 0.4 0.5
-1
(a) Diluon rate D [h ]

30 6 20

30

rO2 [mmol⋅g-1⋅h-1]
25
15 25
rE[mmol⋅g-1⋅h-1]

20 4
rS [g⋅g-1⋅h-1]

20
15 10
15
RQ

10 2
10
5 rCO2
5 5

0 0 0 0
0.0 0.1 0.2 0.3 0.4 0.5
-1
(b) Diluon rate D [h ]

Fig. 12.11: X/D diagram showing yeast production in continuous cultivation. (a) Measurements of
concentrations and volumetric gas turnover rates; (b) calculated specific rates.

go to its maximum before another intracellular step is the bottleneck, e.g., the speed
of DNA replication.
This is easily written down but how is this approach evaluated? To answer this
question we have to take a trip to linear optimization, an approach known as ‘linear
programming’. The kernel of the approach is a linear cost function to be minimized or
maximized. Linear means in principle that it can go to infinity. This is prevented by
linear constraints, given either as linear equations, in the yeast case stoichiometries,
312 | 12 Modeling – art and handcraft of mathematically describing bioprocesses

2 Ethanol
2 CO2
O2

Glucose 6 CO2
TCC 6 H2O

Fig. 12.12: Metabolic scheme of yeast metabolism; as a further flux ethanol production is consid-
ered.

Direcon of
rO2
μ-opmizaon
Maximum oxygen
turnover
Opmal
working point

Aerobic
stoichiometry
Permied
sector Current possible
substrate uptake

rS

Fig. 12.13: Pictorial representation of linear programming for the example of substrate and oxygen
uptake as independent variables. The profit function can be maximized until it finds constraints at
the upper right bound.

or as inequalities to formulate upper or lower limits. A general form for x being the
vector of the ‘adjusting screws’ is:
max (cT × r) subject to (Y × rT ≤ b) and (r ≥ 0) (12.9)
So also a linear combination of unknowns is possible to be chosen as lowest cost or
maximum profit function. The parameter b was 0 in our examples, but could con-
tain other fixed terms like maintenance energy. The situation is visualized for two un-
known variables in Figure 12.13.
Now comes the interesting point from theory: All constraints together span (or
have to span) a multidimensional polygon. The maximum is always reached at ex-
actly one edge of this polygon and not somewhere in the middle. Optimization means
 12.6 Back to baker’s yeast – more degrees of freedom | 313

18 35

Substrate [g·L-1]
16
30
14
25
12
Biomass [g·L-1]

10 20

8 15

Ethanol [g·L-1]
6
10
4
5
2

0 0
0.0 0.1 0.2 0.3 0.4 0.5 0.6

(a) Diluon rate D [h-1]

3.0
rE [g·g-1·h-1]

2.5

2.0

1.5
rS

1.0
rx

0.5

0.0
0.0 0.1 0.2 0.3 0.4 0.5 0.6
(b) Diluon rate D [h-1]

Fig. 12.14: Simulation of a continuous yeast cultivation; parameters are c S,f = 30; M Cat = 90; e N,X =
0.1; rS,max = 2.5; kS = 0.1; y ATP,Cat = 16/M Cat ; y X,ATP = 18/90; y ATP,Eth = 2/90. Concentrations
are plotted in (a), while specific turnover rates are shown in (b). Further details can be taken from
the model in the virtual appendix. Note that the simulator does not deal with units, so consistent
conventions have to be conveyed.

therefore to calculate the linear cost function for all possible edges and find the best
one out of the list of results. In simulation packages there are of course more clever
routines to solve the linear programming problem. Now we have a look at the results
for yeast growth (Figure 12.14).
314 | 12 Modeling – art and handcraft of mathematically describing bioprocesses

The simulation reflects the data quite well (which has to be proven). For low di-
lution rates again the normal situation for aerobic growth is reproduced, especially
the nearly constant biomass concentration. At about D = r X = 0.23 h−1 the specific
oxygen uptake rate reaches its maximum. To grow faster the yeast cells have to pro-
duce ethanol, visible from the increasing ethanol concentration and the increasing
specific ethanol formation rate. The curve for specific substrate uptake increases more
steeply from this point, while biomass concentration sharply drops. This is debited to
the lower ATP yield of the fermentative pathway. Finally, at about r X = 0.42 h−1 the
specific uptake rate also reaches its maximum. Further increase of D is not possible
and the simulation shows the washout case. However, the real yeast data show a dif-
ference. At high dilution rates the specific oxygen uptake rate even decreases. This
worsens the Crabtree effect further, which is of course not good for the yield. We can
at this point only speculate. Oxygen diffusion could be limited but this is unlikely as
there is no visible reason why that should be the case. Another assumption is that
the cell ‘intentionally’ reduces the formation of enzymes in the respiratory chain. The
background could again be an intracellular optimization where these complex en-
zymes are more ‘expensive’ to produce than kinases for sugar transport, or that the
mitochondria do not have enough membrane space at high growth rates, or that the
expression rate is somehow limited. Note that each cell optimizes its local cost crite-
rion. This is not automatically good for the population. (Is this an analog model for
rise in population of mankind?) Another optimization idea for microorganisms could
be efficiency during growth at strong substrate deprivations.
The optimization approach is also a useful tool in complex cultivation patterns,
where the cells run through different limitations. Until now substrate uptake was in-
sinuated as limiting. In cases where only one limiting step is expected, the limitation
could switch between substrate, nitrogen, or oxygen limitation during the process.
Setting all three kinetics as possible constraints for the optimization approach intrin-
sically delivers the switching point between the possible physiological states. Under
limitation conditions like nitrate or phosphate limitation (no protein and nucleic acid
formation is possible) some microorganisms do not reduce substrate uptake or pro-
duce byproducts to fulfill the stoichiometric requirements, but they produce intracel-
lular storage compounds. Examples are PHB production in bacteria or starch and oil
accumulation in microalgae (see Figure 8.26). In such cases the optimization approach
is also useful to simulate this behavior as the simulation examples (virtual material)
prove. In such cases cells may change their physiological behavior. Considering this in
representing different physiological cell types leads to a so called segregated model.
An example for yeast growth, where yeasts in the different cell cycle states play a role,
is given in the virtual material.
12.7 Back to microalgae – a spatial and hierarchical structure for subsystem definition | 315

12.7 Back to microalgae – a more spatial and hierarchical

structure for subsystem definition

As already stated above, a process model can be decomposed into single stages, and
the bioreactor itself into three physical phases. The ‘biophase’ consisting of poten-
tially different cells can be subdivided into submodels for different cell types leading
to segregated models (different cell types from one species) or population models. But
also inside the cells hierarchical structuring is applicable. A cell is not an ideally mixed
reactor. Rather different reactions take place in different part of the cell, often in or-
ganelles as distinct compartments. Models considering this structure are called ‘com-
partment models’. Until now we considered mainly rigid metabolic pathways for cells
with a given composition. Depending on environmental conditions it is entirely pos-
sible that the cells change their macromolecular composition. To do so the cell has a
complex network of sensors, actuators and controllers on the genetic, epigenetic, and
enzymatic level. In Figure 12.15 this is marked as the ‘control level’, one hierarchical
level above the metabolism. It is in principle not possible to model the control level in
detail, so we summarized it in the optimization paradigm. In the following paragraph
an example of a spatial and hierarchical model is elaborated where microalgae and
especially the chloroplasts play the main role.
Photosynthesis located in the chloroplasts consists of light absorption, ATP and
NADH2 generation, and carbon fixation. The dependency of growth and light intensity
and light absorption is condensed in the PI curve (Figure 8.10, Equ. 8.8). Now we have
a closer look at the different steps to distinguish between light absorption and consec-
utive reaction steps. Firstly, the most important specific turnover rates have to be listed

Level of funconal macromolecules

Polysaccharides
Pigments Lipids

?
Acvity Metabolic network Polysaccharides

Pigments
ATP
h·ν Glucose Lipids
NADPH2
Absorpon/ Dark reacon …
light reacon

Reactor model CO2 O2

(Light input, light distribuon, gassing, rheology properes)

Fig. 12.15: Spatially and hierarchically structured graphical model for phototrophic growth. Ellipses
are metabolic compounds, rectangles are material converters (metabolic pathways), rhombs are
intracellular controllers.
316 | 12 Modeling – art and handcraft of mathematically describing bioprocesses

before using them in the model. Light capturing by light absorption, denoted as r hν,abs,
as a passive physical transport step can theoretically go to infinity. The consecutive
steps on photosynthesis have of course a limited capacity, so depending on growth
conditions not all photons can be used. This situation is aggravated as the cells cannot
‘close their eyes’, where a bacterium can downregulate substrate uptake. So we have
to split the photons into used ones r hν,use and radiated ones r hν,NPQ . The energy and re-
dox balance can be set up only for the chloroplasts based on the gross Equation (8.3).
As primary product of photosynthesis starch is accounted for. Formally setting up both
balances would lead to an overdetermined set of equations, as in the anaerobic case.
To balance both, the cell can adjust the relation of redox and energy generation by
shifting between linear and cyclic electron flux. This allows us to lump the process
directly into starch production rStarch from a given amount of photons r hν,use . Further,
CO2 uptake rCO2 is needed as well as oxygen formation rO2 . The final vector of specific
turnover rates in the chloroplasts reads (12.10a):

rPhoto = [r hν,abs r hν,NPQ r hν,use rCO2 rO2 rStarch ] (12.10a)

Setting up the system matrix is now comparatively easy (Equ. 12.10b).

It contains the photon balance in the first row, photon-starch yield in the second
row, and the stoichiometry of starch (here calculated as glucose) from CO2 and O2
formation in the last two rows.

1 −1 −1 0 0 0
[ ]
[0 0 yATP,hν 0 0 −yStarch,ATP ]
YPhoto =[
[0 −1 −1
]
] (12.10b)
[ 0 0 MCO2 0 −6 ⋅ MStarch ]
−1 −1
[0 0 0 MCO2 MO2 0 ]
The consecutive steps on photosynthesis have of course a limited capacity. Maximum
starch production depends on limitations of different reaction steps, either some-
where in water splitting, here noted as r hν,use , or in CO2 availability. This latter one
is predominant in nature. In sum, this leads again to overdetermination, as not all
photons are used. The way out for the cell and for modeling is the release of absorbed
light energy from the light harvesting complexes as photons or heat (NPQ). The opti-
mization approach includes the fact that the cell tries to minimize the wasted photons.
These aspects can now be formulated in a submodel holding for the chloroplasts:
󵄨
max {rStarch 󵄨󵄨󵄨󵄨YPhoto × rTPhoto = 0 ,
r hν,abs = σ X ⋅ I hν ,
(12.10c)
rCO2 ≤ f (cCO2 ) ,
r hν,use ≤ r hν,use,max }

For the heterotrophic part we can assume the same structure as in the aerobic exam-
ple, only that the substrate is the self-produced starch as written in Equation (12.10d).
12.7 Back to microalgae – a spatial and hierarchical structure for subsystem definition | 317

rX,
rStarch
Increasing CO2
Starch limitaon
accumulaon
Increasing NH3
Possible limitaon
growth
rabs,hν

Fig. 12.16: Light response curve (PI curve) for different limitation conditions.

Here, to be on the safe side, maximum growth for other reasons is also considered.

max {r X |Yaerob × raerob = 0 ,

rS ≤ rStarch ,
(12.10d)
rNH3 ≤ fkin (cNH3 ) ,
r X ≤ r X,max }

The consequences for the PI curve become clear from this model reflections and are
visualized in Figure 12.16. For low light intensities the linear branch depends only on
absorbed photons, from which all are used for growth. No additional limitation oc-
curs. The saturation branch however is the consequence of different limitations and
not necessarily from the light reaction. For carbon limitation inside the chloroplast,
starch production is reduced, having direct impact on growth. This is actually a mani-
festation of Blackman kinetics (Chapter 4). NPQ should rise under this condition. Fur-
ther down in the anabolism nitrogen limitation may occur. In fact, we did not consider
any feedback to photosynthesis. The consequence is then an ongoing starch produc-
tion. That is actually the case in many microalgae, where nitrogen deprivation leads
to starch or to lipid accumulation. A simulation example is available in Figure 8.26.
For quantitative analysis the P curve has now to be applied to photobioreactors,
where no homogeneous light conditions prevail. Reactor equations are not simple
mass balances as in the ideally mixed heterotrophic case. The specific growth rate
μ might be constant at μmax in the bright front of the reactor. At an a priori not ex-
actly known point inside the reactor it will decay according to local light intensity and
kinetics. As a comparatively simple case a flat cuboid geometry is chosen, where an
exponential decay of light intensity is observed along the light path.
As we practically cannot measure μ along the light path a formal deduction of
the macroscopically apparent average growth rate μav (I hν,0 ) has to be developed. The
most direct case is to virtually divide the total light path length, here the thickness or
depth of the reactor DR , into infinitesimal slices dlpath and to sum up the respective
local μ(I hν ) over all slices. This is to integrate μ(I hν ) along the light path from the side
of incident light to the dark remote side of the reactor as formally deduced in Equa-
318 | 12 Modeling – art and handcraft of mathematically describing bioprocesses

tion (12.11):
DR
1
μ av,μ (I hν,0) = ∫ μ (I hν (lpath )) ⋅ dlpath (12.11)
DR
0
This approach is called ‘μ integration’, indicated here with the index μ. In fact, a for-
mal integration is not possible for all possible kinetics. Some examples are given in the
supplementary material. A simulation for different σ X comparing ideal local kinetics
μ(I hν ) with observations μav is plotted in Figure 8.19 as a function of the incident light
intensity I hν,0 . Also given is μ hν,I (I hν ) assuming that the cells can somehow store the
absorbed light and grow finally according to the mean value of the light distribution
in the reactor as denoted in Equation (12.12):
DR
1
μ av,I (I hν,0 ) = μ ( ∫ I hν (lpath ) ⋅ dlpath ) (12.12)
DR
0

This approach is called ‘light integration’. Simulation results for the resulting μ versus
I0 curves are simulated in Figure 12.17.
In the case where all cells experience this light, we observe the ideal kinetics.
As expected, light integration leads to a linear increase not reaching the maximum
growth rate. For μ integration, with increasing I0 an increasing part of the reactor is
bright, leading to light saturation and therefore not to higher growth rates. This looks
very similar to Monod kinetics and is therefore sometimes misinterpreted. Real values
can lie a bit higher as microalgae can indeed store activated states for some ns and
ATP level for some ms. Fast mixing with fast changes between dark and light parts of

1.2
ideal kinecs

1.0

I-integraon
0.8
μ [h-1]

μ-integraon
0.6

0.4

0.2

0.0
0 200 400 600 800
I0 [μE⋅m-2⋅s-1]

Fig. 12.17: Average specific growth rate as a function of incident light intensity for different assump-
tions.
 12.8 Integration into society – the final proof of meaning | 319

the reactor from the view of a single cell supports this energy storage leading to the so
called ‘intermittent’ (or ‘flashing’) light effect.
With the optimization approach it is possible to cover some aspects of chang-
ing macromolecular composition of a cell, especially accumulation of storage com-
pounds. Closing up modeling considerations, we now collect some examples for visi-
ble impact of the control level onto flux distribution and macromolecular composition.
Reduction of maximum respiratory capacity in yeast with increasing glucose availabil-
ity is already the first example on the enzymatic level. Having the choice between dif-
ferent substrates like glucose and fructose many species can downregulate the uptake
of all but one. High temperature induces chaperone production in many bacteria, a be-
havior exploited in recombinant protein production. Microalgae avoid losing photons
by NPQ, so most of them are able to reduce chlorophyll content under high radiation
conditions with feedback to flux distribution. With polysaccharides microalgae influ-
ence their environment in a controlled manner. While flux models for bioprocesses
are usually defined as stationary, control reactions on the level of macromolecules
show time constants in the range of minutes or hours. So dynamic modeling has to be
applied in these cases. Make notes about other cases which you get to know!
Even in a late stage of development bioprocess models contain only weakly known
relations or unknown parameters. These should be addressed by model based exper-
imental design. Here is finally in this chapter a saying of Norbert Wiener (who coined
the term ‘cybernetics’): “The best material model of a cat is another, or preferably the
same, cat.”

12.8 Integration into society – the final proof of meaning

All the attempts to develop bioprocesses are only sensible if somebody needs and
wants the product. That sounds self-evident but has different characteristics in differ-
ent fields of application. For medical applications cost benefit ratio is rated differently
as for food, cosmetics, or bulk chemicals. Typically, competing chemical products can
be found in the latter case, while therapeutics are often natural or nature identical
substances, where microorganisms have huge advantages as producers. Not only eco-
nomic aspects play a role, but in addition and in mutual dependency ecological, ethi-
cal, societal, or simply practical aspects can be decisive for acceptance in the market.
This also affects the process as such. Use of renewable resources and the avoidance of
organic solvents are critical points. Labels like ‘bio’, ‘without mineral oils’ or ‘without
animal additives’ count in the market. Things like diatomaceous earth as filter aid are
not in public awareness but are worth considering from the ecological view as well.
It should not be forgotten that there are people running the process in a factory. Pro-
cess design to allow daily or weekly production cycles mark only one consequence.
Worker skill and education is a concern in process development. Finding and paying
a bioengineer could be a knockout criterion for a sophisticated bioprocess with mi-
320 | 12 Modeling – art and handcraft of mathematically describing bioprocesses

Tab. 12.1: Aspects of integration on the societal level.

Field Aspect Measures and examples

Direct impact on humans Healthcare Antibodies (mammalian cell
culture), recombinant proteins
(E. coli)
Food processing, food supply Fermentation (Lactobacillus),
microalgae (Chlorella)
Cosmetics, wellness
Military use ‘The Dirty Dozen’
Working world Occupational safety, product Reactor design, GMP, PAT
safety
Work rhythm, shift work Process duration
Existing infrastructure Biorefinery, closed material cycles
Education and training
Environment Environmental protection Wastewater treatment
Renewable resources No organic solvents
Biodegradable products Polylactic acid
Economics Production costs Process intensification
Market price Specific advertising
Political, cultural and ethical Public perception Gene technology, animal welfare
aspects
National regulations ‘Bio’ label
Tradition Cheese
Sciences Integration of sciences Extensive networking, education
Art and design Architecture

croorganisms, especially for small companies. The mentioned aspects, together with
the amount of wages, are criteria for the decision of in which country a production
process is established. In the previous chapters some examples have been outlined
for different aspects of integration into society. These are ordered in Table 12.1 not for
the sake of completeness but to encourage reconsidering of the examples in this book
and making your own thoughts. As bioengineers we are called to integrate ourselves
in scientific networks and the fast developing vertically and horizontally structured
scientific communities.
What will be the future of bioprocess engineering? As the future is per se not pre-
dictable, asking questions about current trends may be sufficient here. Will synthetic
biology provide standard organisms for all products or even artificial cells making pro-
cess development superfluous? Modern enzyme processes with enzyme cascades and
cofactor regeneration could make the employment of living cells obsolete as well. Will
better understanding of biology make scale up from microtiter plates to production
scale possible, again with less engineering input? Or is it imaginable the other way
around: that more and more engineering thinking is needed to understand intracellu-
lar thermodynamic processes? Further interlocking between biology and bioprocess
 12.8 Integration into society – the final proof of meaning | 321

engineering is mandatory. Food processing is traditionally the field of most interfer-

ence between technical production process and public perception. With respect to bio-
engineering, fermented food with genetically engineered microorganisms is an exam-
ple, or the newest developments in making meat substitutes with animal cell cultures.
The science fiction author Isaac Asimov dared a prediction at Expo 1964: “Ordinary
agriculture will keep up with great difficulty and there will be ‘farms’ turning to the
more efficient microorganisms. Processed yeast and algae products will be available in
a variety of flavors . . . but there will be considerable psychological resistance to such
an innovation.” This statement could be made in these days in the same way. Scien-
tific targets changed with changing economic constraints, in this case from making
yeast from oil to making oil from yeast. In addition, taste, smell and color are still
improvable. Raw material from microorganisms for food production is a small sector;
microalgae could make the breakthrough.
A desirable trend is to capture more and more areas. Since the wonderful drawings
of diatoms in the famous book of Ernst Haeckel ‘Kunstformen der Natur’ (Art Forms
in Nature, 1904), microorganisms have been taken as motifs for visual arts. In modern
form they can even been found as cartoons or living bacterial colonies on Petri dishes.
An example for using microorganisms in art is shown in Figure 12.18. The artist Otto
Roth applied living microalgae on an illuminated glass plate and captured the lumi-
nescence with a camera. Nevertheless, bioprocesses as such have found, despite their
societal importance and sometimes controversial discussion, only little artistic im-
pact. Has somebody an idea for a picture or painting to show the pros and cons of
bioethanol production?

Fig. 12.18: Light art canopy: ‘Meeresleuchten’ (milky seas effect) Plankton luminogram of Pyrocystis
elegans, technicolor, 8×10 cm, 2007. Tim Otto Roth in Cooperation with Prof. Dr. Rüdiger Hardeland,
Germany. © Tim Otto Roth, imachination projects.
322 | 12 Modeling – art and handcraft of mathematically describing bioprocesses

Solar panel Algae fountain Tube system

Bent wood structure

Indoor light

Flexible indoor space

Algae tank Circulaon system

Fig. 12.19: Example of new nature inspired architecture, Recreation Room Chlorella © Adam Miclosi,
Hungary.

Architects are also inspired by nature. Again, diatoms are the favorite templates to
reach optimum structures between stability and low material efforts. Hanging gardens
with mosses and microalgae filled photobioreactors are hypes in architectural designs
of future cities.
Dear reader, now you have reached the end of this book and deserve a break. The
author recommends the recreation dome of the designer Adam Miclosi (Figure 12.19)
enriched with original Chlorella oxygen!

12.9 Suggestions

1. Add into the metabolic scheme the number of carbons in the metabolic flux
schemes and the number of NADH2 and ATP. Adjust the stoichiometric matri-
ces accordingly. As these are linear operations, the simulation results should
not change.
2. Now go step by step through the simulations in the virtual material and try to
verify the models in this chapter.
3. In the microalgae example there is a serious oversimplification making an
adjustment of the heterotrophic submodel necessary. Starch contributes to
dry weight, but not to light absorption and the growth machinery. Decompose
growth into functional (active) biomass c X,func and intracellular starch.
Further Reading – still curious?
The following list contains additional written material where you can find access to
more information of interest. Some topics require additional background, e.g., micro-
biology, where other text books may turn out to be helpful. Other topics are not covered
by this book to the extent to which they should. This is due to available space but also
to the need to show different principles by examples. Not all the example processes
are supported by references, sometimes simply because there are no good reviews. Be-
cause of the integrative character, the list is intentionally not ordered chapter by chap-
ter. However, it is an invitation to rummage in references, to consolidate and deepen
what has been learned, and above all to get your own creative ideas.

Books and databases

[1] Baltz, R. H.; Davies, J. E.; Demain, A. L. (2010): Manual of industrial microbiology and biotech-
nology. 3rd edn. Washington, DC: ASM Press.
[2] Gupta, K. V. (2016): Microbial Applications. Recent Advancements and Future Develop-
ments. Edited by Zeilinger, S.; Ximenes Ferreira Filho, E.; Durán-Domínguez-de-Bazúa, M.;
Purchase, D.. Berlin, Boston: De Gruyter. Available online at https://doi.org/10.1515/
9783110412789.
[3] Henkel, M.; Zwick, M.; Beuker, J.; Willenbacher, J.; Baumann, S.; Oswald, F. et al. (2015): Teach-
ing bioprocess engineering to undergraduates. Multidisciplinary hands-on training in a one-
week practical course. Biochemistry and molecular biology education: a bimonthly publication
of the International Union of Biochemistry and Molecular Biology 43 (3), pp. 189–202. DOI:
10.1002/bmb.20860.
[4] Ingalls, B. P. (2013): Mathematical modeling in systems biology. An introduction. Cambridge,
Mass.: MIT Press. Available online at http://site.ebrary.com/lib/alltitles/docDetail.action?
docID=10734711.
[5] Kadic, E.; Heindel, T. J. (2014): An Introduction to Bioreactor Hydrodynamics and Gas–Liquid
Mass Transfer. 1st edn. New York: Wiley. Available online at http://search.ebscohost.com/
login.aspx?direct=true&scope=site&db=nlebk&db=nlabk&AN=761798.
[6] Liu, S. (2013): Bioprocess engineering. Kinetics, biosystems, sustainability, and reactor de-
sign. Amsterdam: Elsevier. Available online at http://site.ebrary.com/lib/alltitles/docDetail.
action?docID=10599030.
[7] Madigan, M. T.; Martinko, J. M.; Bender, K. S.; Buckley, D. H.; Stahl, D. A. (2015): Brock biology
of microorganisms. 14th edn., global edn. Boston: Pearson (Always learning).
[8] Schlegel, H. G. (1993): General Microbiology. 7th edn. Cambridge: Cambridge Univ. Press.
Contains all necessary biological background for bioengineers.
[9] Simpson, R.; Sastry, S. K. (2013): Chemical and Bioprocess Engineering. Fundamental Con-
cepts for First–Year Students. New York, NY, s.l: Springer New York. Available online at
http://dx.doi.org/10.1007/978-1-4614-9126-2.
[10] Zorn, H.; Czermak, P.; Langer, U. (2009): Improving bioreactor cultivation conditions for sensi-
tive cell lines by dynamic membrane aeration, Cytotechnology, 53, pp. 17-30.

https://doi.org/10.1515/9783110315394-013
324 | Further Reading – still curious?

[11] BioNumbers –The Database of Useful Biological Numbers. Available online at http://
bionumbers.hms.harvard.edu/, checked on 12/8/2017. An amazing abundance of quantita-
tive information.
[12] Enzyme Database – BRENDA. Available online at http://www.brenda-enzymes.org/index.php,
checked on 8/25/2016.
[13] Richards, M. A.; Cassen, V.; Heavner, B. D.; Ajami, N. E.; Herrmann, A.; Simeonidis, E.;
Price, N. D. (2014): MediaDB. A database of microbial growth conditions in defined media.
PLoS ONE 9 (8), e103548. DOI: 10.1371/journal.pone.0103548.
[14] Science and medical royalty free and rights managed images, photos, illustrations – Sci-
ence Photo Library. Available online at http://www.sciencephoto.com/images, checked on
12/8/2017.
[15] When we share, everyone wins – Creative Commons. Available online at https://
creativecommons.org/, checked on 12/8/2017.

Reviews and good papers on technical aspects

and cross-sectoral issues

[1] Adrio, J.-L.; Demain, A. L. (2010): Recombinant organisms for production of industrial products.
Bioengineered Bugs 1 (2), pp. 116–131. DOI: 10.4161/bbug.1.2.10484.
[2] Biechele, P.; Busse, C.; Solle, D.; Scheper, T.; Reardon, K. (2015): Sensor systems
for bioprocess monitoring. Engineering in Life Sciences 15 (5), pp. 469–488. DOI:
10.1002/elsc.201500014.
[3] Chisti, Y.; Jauregui-Haza, U. J. (2002): Oxygen transfer and mixing in mechanically agitated
airlift bioreactors. Biochemical Engineering Journal 10 (2), pp. 143–153.
[4] Chisti, Y.; Moo-Young, M. (2013): Bioreactors. In: Encyclopedia of Physical Science and Technol-
ogy. 3rd edn. Amsterdam, Elsevier Science, pp. 247–271.
[5] Ehrenberg, M.; Bremer, H.; Dennis, P. P. (2013): Medium-dependent control of the bacterial
growth rate. Biochimie 95 (4), pp. 643–658. DOI: 10.1016/j.biochi.2012.11.012.
[6] Eibl, D.; Eibl, R.; Adams, T. (2014): Disposable bioreactors. Advances in Biochemical Engineer-
ing/Biotechnology 138, pp. 1–44.
[7] Frahm, B.; Brod, H.; Langer, U. (2009): Improving bioreactor cultivation conditions for sen-
sitive cell lines by dynamic membrane aeration. Cytotechnology 59 (1), pp. 17–30. DOI:
10.1007/s10616-009-9189-9.
[8] Glassey, J. (2013): Multivariate Data Analysis for Advancing the Interpretation of Bioprocess
Measurement and Monitoring Data. Measurement, Monitoring, Modelling and Control of Bio-
processes 132, pp. 167–191. DOI: 10.1007/10_2012_171.
[9] Havlik, I.; Scheper, T.; Reardon, K. F. (2016): Monitoring of Microalgal Processes. Advances in
Biochemical Engineering/Biotechnology 153, pp. 89–142. DOI: 10.1007/10_2015_328.
[10] Kaiser, S. C.; Werner, S.; Jossen, V.; Kraume, M.; Eibl, D. (2017): Development of a method for
reliable power input measurements in conventional and single-use stirred bioreactors at labo-
ratory scale. Engineering in Life Sciences 17 (5), pp. 500–511. DOI: 10.1002/elsc.201600096.
[11] Mandenius, C.-F.; Titchener-Hooker, N. J. (© 2013): Measurement, monitoring, modelling and
control of bioprocesses. Advances in Biochemical Engineering/Biotechnology 132, pp. 1–33
[12] Oberhardt, M. A.; Zarecki, R.; Gronow, S.; Lang, E.; Klenk, H.-P.; Gophna, U.; Ruppin, E. (2015):
Harnessing the landscape of microbial culture media to predict new organism-media pairings.
Nature communications 6, p. 8493. DOI: 10.1038/ncomms9493.
 Publications on specific organisms or processes with integrative character | 325

[13] Schirmer, M.; Posten, C. (2016): Modeling of microalgae bioprocesses. Advances in Chemical
Engineering 48, pp. 151–184.
[14] von Stockar, U.; van der Wielen, L. A. M. (2013): Thermodynamics in Biochemical Engineering.
Hoboken: EFPL Press. Available online at http://gbv.eblib.com/patron/FullRecord.aspx?p=
1402693.
[15] Stouthamer, A. H. (2012): Quantitative Aspects of Growth and Metabolism of Microorganisms.
Amsterdam: Springer Netherlands. Available online at https://books.google.de/books?id=
lgfwCAAAQBAJ.
[16] Suresh, S.; Srivastava, V. C.; Mishra, I. M. (2009): Techniques for oxygen transfer measurement
in bioreactors. A review. Journal of Chemical Technology and Biotechnology 84 (8), pp. 1091–
1103. DOI: 10.1002/jctb.2154.
[17] Villaverde, A. F.; Bongard, S.; Mauch, K.; Balsa-Canto, E.; Banga, J. R. (2016): Metabolic en-
gineering with multi-objective optimization of kinetic models. Journal of Biotechnology 222,
pp. 1–8. DOI: 10.1016/j.jbiotec.2016.01.005.

Publications on specific organisms or processes

with integrative character

[1] Ault, A. (2004): The Monosodium Glutamate Story. The Commercial Production of MSG and
Other Amino Acids. J. Chem. Educ. 81 (3), p. 347. DOI: 10.1021/ed081p347.
[2] Béchet, Q.; Chambonnière, P.; Shilton, A.; Guizard, G.; Guieysse, B. (2015): Algal productivity
modeling: A step toward accurate assessments of full-scale algal cultivation. Biotechnology
and Bioengineering 112 (5), pp. 987–996. DOI: 10.1002/bit.25517.
[3] Borchert, S.-O.; Voss, T.; Schuetzmeier, F.; Paul, J.; Cornelissen, G.; Luttmann, R. (2015):
Development and monitoring of an integrated bioprocess for production of a potential
malaria vaccine with Pichia pastoris. Journal of Process Control 35, pp. 113–126. DOI:
10.1016/j.jprocont.2015.08.006.
[4] Dillschneider, R.; Posten, C. (2013): Closed bioreactors as tools for microalgae production. In:
Advanced Biofuels and Bioproducts. New York, Springer, pp. 629–649.
[5] Elander, R. P. (2003): Industrial production of beta-lactam antibiotics. Applied Microbiology
and Biotechnology 61 (5–6), pp. 385–392. DOI: 10.1007/s00253-003-1274-y.
[6] Frahm, B.; Brod, H.; Langer, U. (2009): Improving bioreactor cultivation conditions for sen-
sitive cell lines by dynamic membrane aeration. Cytotechnology 59 (1), pp. 17–30. DOI:
10.1007/s10616-009-9189-9.
[7] Fresewinkel, M.; Rosello, R.; Wilhelm, C.; Kruse, O.; Hankamer, B.; Posten, C. (2014): Integra-
tion in microalgal bioprocess development. Engineering in Life Sciences 14 (6), pp. 560–573.
DOI: 10.1002/elsc.201300153.
[8] Kandilian, R.; Soulies, A.; Pruvost, J.; Rousseau, B.; Legrand, J.; Pilon, L. (2016): Simple
method for measuring the spectral absorption cross-section of microalgae. Chemical Engi-
neering Science 146, pp. 357–368. DOI: 10.1016/j.ces.2016.02.039.
[9] Mattanovich, D.; Jungo, C.; Wenger, J.; Dabros, M.; Maurer, M. (2014): Yeast suspension cul-
ture. In: Industrial Scale Suspension Culture of Living Cells. Weinheim, Wiley-VCH, pp. 94–129.
[10] Ozcengiz, G.; Demain, A. L. (2013): Recent advances in the biosynthesis of penicillins,
cephalosporins and clavams and its regulation. Biotechnology Advances 31 (2), pp. 287–311.
DOI: 10.1016/j.biotechadv.2012.12.001.
326 | Further Reading – still curious?

[11] Pappenberger, G.; Hohmann, H.-P. (2014): Industrial production of L-ascorbic Acid (vitamin C)
and D-isoascorbic acid. Advances in Biochemical Engineering/Biotechnology 143, pp. 143–
188. DOI: 10.1007/10_2013_243.
[12] Reed, G.; Nagodawithana, T. W. (1990): Yeast Technology. 2nd edn. Dordrecht: Springer Nether-
lands. Available online at http://dx.doi.org/10.1007/978-94-011-9771-7.
[13] Safi, C.; Zebib, B.; Merah, O.; Pontalier, P.-Y.; Vaca-Garcia, C. (2014): Morphology, composi-
tion, production, processing and applications of Chlorella vulgaris. A review. Renewable and
Sustainable Energy Reviews 35, pp. 265–278. DOI: 10.1016/j.rser.2014.04.007.
[14] Schippers, A.; Hedrich, S.; Vasters, J.; Drobe, M.; Sand, W.; Willscher, S. (2014): Biomin-
ing. Metal recovery from ores with microorganisms. Advances in Biochemical Engineering/
Biotechnology 141, pp. 1–47. DOI: 10.1007/10_2013_216.
[15] Schmideder, A.; Hensler, S.; Lang, M.; Stratmann, A.; Giesecke, U.; Weuster-Botz, D. (2016):
High-cell-density cultivation and recombinant protein production with Komagataella pastoris
in stirred-tank bioreactors from milliliter to cubic meter scale. Process Biochemistry 51 (2),
pp. 177–184. DOI: 10.1016/j.procbio.2015.11.024.
[16] Schmideder, A.; Weuster-Botz, D. (2017): High-performance recombinant protein production
with Escherichia coli in continuously operated cascades of stirred-tank reactors. Journal of
Industrial Microbiology and Biotechnology, pp. 1–9. DOI: 10.1007/s10295-017-1927-y.
[17] Soundrarajan, N.; Cho, H.-S.; Ahn, B.; Choi, M.; Le Thong, M.; Choi, H. et al. (2016): Green fluo-
rescent protein as a scaffold for high efficiency production of functional bacteriotoxic proteins
in Escherichia coli. Scientific Reports 6, p. 20661. DOI: 10.1038/srep20661.
[18] Tredici, M. R.; Bassi, N.; Prussi, M.; Biondi, N.; Rodolfi, L.; Chini Zittelli, G.; Sampietro, G.
(2015): Energy balance of algal biomass production in a 1-ha “Green Wall Panel” plant. How
to produce algal biomass in a closed reactor achieving a high Net Energy Ratio. Applied Energy
154, pp. 1103–1111. DOI: 10.1016/j.apenergy.2015.01.086.
[19] Weiland, P. (2010): Biogas production. Current state and perspectives. Applied Microbiology
and Biotechnology 85 (4), pp. 849–860. DOI: 10.1007/s00253-009-2246-7.
[20] Wolf, J.; Stephens, E.; Steinbusch, S.; Yarnold, J.; Ross, I. L.; Steinweg, C., et al. (2016): Multi-
factorial comparison of photobioreactor geometries in parallel microalgae cultivations. Algal
Research 15, pp. 187–201. DOI: 10.1016/j.algal.2016.02.018.
[21] Zhang, J.; Greasham, R. (1999): Chemically defined media for commercial fermentations. Ap-
plied Microbiology and Biotechnology 51 (4), pp. 407–421. DOI: 10.1007/s002530051411.
[22] Zittelli, G. C.; Biondi, N.; Rodolfi, L.; Tredici, M. R. (2013): Photobioreactors for Mass Production
of Microalgae. In: Handbook of Microalgal Culture: Applied Phycology and Biotechnology. New
York, John Wiley & Sons, pp. 225–266.
Acknowledgments –
dedicated to all the people who supported the
compilation of this book
Writing this book was mainly fun. It helped to clear up my own thoughts on scientific
aspects and to think about how best to convey them to the reader by text, tables, pic-
tures and (sorry to say) mathematical formulas. But it also meant quite some effort.
This concerned many practical things, editorial work, and data retrieval. But profes-
sional scientific support also turned out to be necessary, in order to fill the different
topics with living information. This was only possible with the help of many people,
who supported me with tireless work and creative ideas. It is not possible to express
all the shades of gratitude I feel for those people, and certainly not to weigh their con-
tributions. But I feel the need to let the reader know who it was and what he or she
did, thus giving a little bit back. Please find below some small remarks to them, their
names listed in alphabetical order.

My coworker Dr. Meike Dössel (KIT, Germany) is chemist and

contributed by drawing chemical structures and providing
expert advice on the medium chapter.

My former coworker Prof. Dr. Gözde Gözke (University of

Yalova, Turkey) gave substantial support in setting up the
reactor chapter.

https://doi.org/10.1515/9783110315394-014
328 | Acknowledgments – dedicated to all the people who supported the compilation of this book

Among the students, Eva Heymann (student of bioengineer-

ing, KIT) is noted for her photographic skills: all the reactor
and product pictures are her work.

The specialty of my coworker Dr. Ioanna Jakob (KIT) is bio-

processes development, where her knowledge on inorganic
particle formation was useful in the continuous cultivation
chapter.

Mirco Katzenmeyer (KIT) is a coworker with a chemical engi-

neering background; he delivered flow sheets and other in-
puts to the bioreactor and the fermentation chapters.

Dr. Viktor Klassen (Bielefeld University, Germany) is an ex-

pert in microbiology, especially microalgae in biogas produc-
tion. His huge contribution in the form of pictures and text
blocks was indispensable for the biosystems chapter and
the according paragraphs in the continuous culture and mi-
croalgae chapters.
Acknowledgments – dedicated to all the people who supported the compilation of this book | 329

My secretary Lisa Rest (KIT) cared for the correct sequence of

figure and table numbers and in general for all formal issues
of the manuscript. Further, she invented the nice generic
product labels.

My former lab engineer Lena von Riesen (KIT) carried out

bioreactor runs to gather native data and pictures.

Dr. Rosa Rosello (KIT) was something like a general editor

of the writing process, from formal things to logical progres-
sion.

I know Jens Rupprecht (Sartorius, Germany) from a former

project in Australia. Now he is involved in reactor develop-
ment and shared his insight by discussing current trends in
bioreactor development with me.
330 | Acknowledgments – dedicated to all the people who supported the compilation of this book

An immense amount of work and ideas came from Selima

Sayah (student of bioengineering, KIT) by creating most of
the pictures in PowerPoint, handling all the data plots, and
giving feedback from the perspective of a student.

The enlightening simulations of the sun owe their existence

to Kira Schediwy (KIT), coworker at my institute.

The support of the institute in all engineering questions was

Christian Steinweg (bioprocess engineer, KIT); he delivered
the CAD reactor drawings.

Renate Suda (IIT, Germany) is bioscientific documentalist.

She performed literature retrieval including more specific is-
sues.
Acknowledgments – dedicated to all the people who supported the compilation of this book | 331

Michael Taebling (alias Käsemichl, Schönegger Käsealm,

Germany) brought to me the idea of cheese making as a sig-
nificant bioprocess and shared with me some secrets about
it.

Dr. Andreas Trautmann (KIT, Lonza, Switzerland) super-

vised with great dedication the fermentation course of my
institute and condensed his experiences in the fermentation
chapter.

Dr. Ines Wagner (KIT, OHB, Germany) worked as scientific

coworker in my group and made her PhD in the field of mi-
croalgae; the photosynthesis part in this book is a highly wel-
come spinoff.

My former college Prof. Dr. Şems Yonsel (Okan University,

Turkey) worked several years in a baker’s yeast factory and
knows the process from the theoretical and the practical
side, making him a perfect advisor for the running yeast ex-
ample.
332 | Acknowledgments – dedicated to all the people who supported the compilation of this book

Special thanks goes also to Marc Bisshops (Director SLS Biopharm, Pall Life Sciences),
who discussed the pros and cons of continuous processing and wrote the precise state-
ment in the chapter.
Last but not least I want to say thank you to Ria Fritz and Vivien Schubert from
the publisher De Gruyter for their encouragement to publish this book, practical help
and advice with the manuscript, and their professional patience when the delivery of
the manuscript was again delayed.

To express my sincere gratitude: THANK YOU SO MUCH!!!

Clemens Posten
Copyrights – pictures provided with courtesy
and accepted with thanks
The following persons and companies delivered pictures and kindly provided copy-
right for them to be printed in this book:
Fig. 1.1 a: Meine Pestoria, Grötzinger Straße 42–44, 76227 Karlsruhe, Germany.
Fig. 1.1 b: Schönegger Käse-Alm GmbH, Steinwies 20, 86984 Prem, Germany, with special
thanks to Matthias Köpf and Michael Taebling (Michel’s Rollender Käseladen,
Eggweg6, 86989 Steingaden).
Fig. 2.2 a, b: Olga Blifernez-Klassen, personal communication by Viktor Klassen.
Fig. 2.3 a: From Hutchison, Clyde A. et al., ‘Design and synthesis of a minimal bacterial
genome’. In: Science (New York, N. Y.) 351 (6280), aad6253. DOI: 10.1126/sci-
ence.aad6253. Reprinted with permission from AAAS.
License Number: 3953251395268; license bought by Viktor Klassen.
Fig. 2.3 b: Courtesy of CDC/ National Escherichia, Shigella, Vibrio Reference Unit at CDC.
Fig. 2.4 a: Viktor Klassen, KIT; culture provided by Jan-Philipp Schwarzhanz, University of
Bielefeld.
Fig. 2.4 b: Permission achieved by communication (Research Gate) from book chapter
“Sample Preparations for Scanning Electron Microscopy – Life Sciences” both
authors: Patchamuthu Ramasamy, Mogana Das, Courtesy of EM Unit, University
Sains, Malaysia.
Fig. 2.5 a: Permission achieved by communication (e-mail) with author Marian Petre,
taken from “Environmental Biotechnology for Bioconversion of Agricultural and
Forestry Wastes into Nutritive Biomass”. In: Marian Petre (Hg.): Environmental
Biotechnology – New Approaches and Prospective Applications: InTech.
Fig. 2.5 b: Permission achieved by communication (e-mail) with authors Ms. Wardah, Mo-
gana Das Murtey, Patchamuthu Ramasamy, courtesy of Ms. Wardah, taken from
“Sample Preparations for Scanning Electron Microscopy – Life Sciences. Modern
Electron Microscopy in Physical and Life Sciences.” InTech.
Fig. 2.6 a: Source: Viktor Klassen, KIT, University Bielefeld, with many thanks for the efforts.
Fig. 2.6 b: Courtesy of Steve Groom, Plymouth Marine Laboratory.
Fig. 2.7 b: Courtesy of Juliane Steingröver, TU-Dresden, Germany.
Fig. 2.8 a: Courtesy of Eric Gottwald, KIT-IFG.
Fig. 2.8 b: Courtesy of Iznewton (own work) [CC BY-SA 4.0 (http://creativecommons.org/
licenses/by-sa/4.0)], via Wikimedia Commons.
Fig. 2.9 a: Courtesy of Michael Wagner, KIT-IFG.
Fig. 2.9 b: Courtesy of Ursula ‘Uschi’ Obst, KIT-IFG, redrawn and modified by authors.
Fig. 5.13: Redrawn from material of CyBioTec MicroReactor.
Fig. 5.14: With kind permission of Eppendorf AG, Bioprocess Center Jülich, Germany, com-
municated by Ulrike Becken.
Fig. 5.15: General Electric company, GE Healthcare, reproduced by permission of the
owner, with thanks to Yana Ostrovskiy and Stijn Vercammen.
Fig. 5.16: With kind permission of Sartorius-Stedim, Germany, communicated by Jens Rup-
precht.

https://doi.org/10.1515/9783110315394-015
334 | Copyrights – pictures provided with courtesy and accepted with thanks

Fig. 5.17: With kind permission of Infraserv-Hoechst, Frankfurt am Main, Germany, com-
municated by Jasmin Graf.
Fig. 5.23: With kind permission of the authors/developers Björn Frahm, University of
Applied Sciences, Ostwestfalen-Lippe and Helmut Brod, Bayer Technology Ser-
vices, Leverkusen, Germany. Described in “Improving bioreactor cultivation
conditions for sensitive cells by dynamic membrane aeration“ Cytotechnology,
2009, 59, 17–30.
Fig. 6.6: Picture and permission achieved from GEA Wiegand GmbH, communicated by
Norbert Strieder with many thanks for the interesting discussion.
Fig. 6.11: Permission achieved by communication (e-mail) with author Jian Ju, thanks
to her, taken from the book chapter “Generation and Utilization of Microbial
Biomass Hydrolysates in Recovery and Production of Poly-(3-hydroxybutyrate),
InTech”.
Figs. 6.13, 6.14: Courtesy of Matthias Köpf, Schönegger Käse-Alm GmbH, Steinwies 20, 86984
Prem, Germany.
Fig. 7.14: Permission achieved by grateful communication (e-mail) with author Chankyu
Park, taken from “Green fluorescent protein as a scaffold for high efficiency pro-
duction of functional bacteriotoxic proteins in Escherichia coli, Scientific Re-
ports, NP”.
Fig. 8.1: Picture taken and provided by Viktor Klassen.
Figs. 8.1 b, 8.2 a, b: Licensed from Science Photo, StockFood GmbH, Munich, Germany, communi-
cated by Martina Braun, with special thanks to Mike Allen, Plymouth Marine Lab-
oratory, UK, who provided some of the pictures.
Figs. 8.14 a, 8.20: Taken at the plant of algae for future a4f in Pataias, Portugal, communicated by
Edgar Santos, greetings to Vitor Verdelho Vieira.
Fig. 8.15: Taken at TUM-AlgaeTec Center, Ottobrunn, Germany, provided by Dirk Weutsr-
Botz, Technical University Munich, Germany.
Fig. 8.18: Taken at the algae pad of KIT by Mirco Katzenmeyer, Germany.
Fig. 8.22: With kind permission from Mira Karst, almostec/bio-compete, Vienna, Austria.
Thanks for this nice idea.
Fig. 9.11: Taken by Martina Nolte, Creative Commons CC-by-sa-3.0 de
(http://creativecommons.org/licenses/by-sa/3.0/de/legalcode),
communicated by Viktor Klassen.
Fig. 9.12: Redrawn by Axel Schippers, Federal Institute for Geosciences and Natural Re-
sources, Hanover, Germany after Johnson et al., Curr. Opin. Biotechnol., 2014.
Fig. 9.13: Licensed by Science Photo.
Fig. 9.15 a: REM picture taken by Nikolay Krumov, KIT, Lonza.
Fig. 9.15: ESEM picture taken by Frank Friedrich, KIT.
Fig. 11.1: From printed user material of bioengineering bioreactor with permission of Bio-
engineering AG, Wald, Switzerland.
Figs. 11.4, 11.5: Taken at KIT owned bioreactor, with permission of Bioengineering AG, 8636
Wald, Switzerland, communicated by Claudia Kälin.
Fig. 12.18: With courtesy of Tim Otto Roth in cooperation with Prof. Dr. Rüdiger Hardeland,
imachination projects, Oppenau, Germany, contact Miriam Seidler.
Fig. 12.19: With kind permission of Adam Miklosi, Hungary, who rendered this drawing es-
pecially for this book, thanks for that.
Index

A anoxia 240
absorbance 199, 264 anoxic milieu 231
absorption 263 antibiotics 4, 41
acetaldehyde 117 antifoam
acetate 2, 145, 150, 176, 197, 238 – agents 93
Acetic acid see acetate – sensor 282
acetoclastic methanogens 233 antioxidant 211
acetogenesis 231 applications 24, 25
acidification 148 aquaculture 213
acidogenesis 231 aqueous solution 152
acidophilic metal sulfide 239 Arrhenius equation 34
activator-inhibitor systems 301 Arthrospira platensis see Spirulina
active substances 29 artificial fertilizers 238
activity of enzymes 19 Aspergillus 2, 52, 170
Adenosine triphosphate (ATP) 132–134, 189, Aspergillus niger 150
304 at-line measurements 261
aerobic ATP/NADH2 ratio 308
– cultivation 117 autoclaving 277, 279, 284, 291
– fed-batch process 272 autoclaving process 281
– growth 44 automated sampling system 252
– growth on ethanol 293 auxotrophic 57
– growth on glucose 132 axenic 110, 277
– heterotrophs 237 axenic culture 94, 195
– metabolism 303
– microorganisms 85 B
– processes 34 Bacillus 176
air inlet device 279 Bacillus subtilis 174
airlift bioreactors 106 bacteria 96, 165, 174, 273
alcohol dehydrogenase (ADH) 293 – species 21
alcohol oxidases (AOX) 179 baffle 95, 106, 280
amino acids 2, 266 Baker’s yeast 8, 24, 161, 293, see also yeast
ammonia 116, 135, 154, 164, 180 balance equations 304
– inhibition 234 balanced growth 46
anabolism 85, 133, 317 batch
anaerobic – cultivations 134
– cultivation 138 – experiments 44
– digestion processes 231 – phase 176, 179, 276
– fermentation 137, 140, 150, 226 – process 86, 146, 148, 153, 201
– growth 85, 145 Beer–Lambert law 199
– metabolism 34, 109 bioactive capabilities 174
– processes 34, 109 bioavailability 212
anaerobic digestion 24 biocatalysis 19
anaplerotic reactions 304 biocatalysts 61, 229
animal cell cultures 95 biodiesel 57
animal cells 113 bioengineer see bioprocess engineer

https://doi.org/10.1515/9783110315394-016
336 | Index

bioengineering 13 cell dry mass 10, 79

bioethanol 7, 138 cell membrane structure 47
biofilm 33, 108, 109 centrifugal field 165
biofuels 29, 226 centrifugation 165, 185
biogas plants 231 chelate agent 150, 195
bioleaching 238 chemical engineering 262, 297
biomass 34, 222, 262, 288 chemolithoautotrophs 237
– concentration 161 chemolithotrophic
biomineralization process 242 – oxidation 238
biopharmaceuticals 31 chemolithotrophs 237
bioprocess engineer 7 chemostat 224, 226
bioprocess engineering 25, 40, 174, 220, 252, Chlamydomonas reinhardtii 186, 187
297, 320 Chlorella vulgaris 185
bioprocesses 18, 83, 92, 111, 117, 153 Chlorophyceae 185
– designing 8, 66 chlorophylls 189, 196, 214
bioreaction stage 11 chloroplasts 125, 183, 187
bioreactor 9, 92, 94, 250, 272 Chlostridium 232
biorefinery 320 chromatography 269, 289
biosensors 269, 270 chymosin 148
Bioweapons 63 citric acid 2, 110, 145, 150, 151, 293
bleeding stream 229 Clark-electrode 259
bottleneck 144, 150, 162, 198, 296, 311 classification 20, 33
bubble column bioreactors 104, 105 cleaning in place (CIP) 94, 141, 291
budding 24 closed photobioreactor 206
budding index (BI) 162 coccoliths 242
buoyancy forces 108, 165 coenzyme 53, 76
butanol 238 cofactors 20, 47, 58
by-pass 251, 261 combined heat and power CHP 235
common iron effect 52
C community 32
C13-labeled substrate 308 complete oxidation of glucose 305
calcite 242 complex molecules 14
calibration curve 289 composter 109
carbon dioxide production rate (CPR) 117 concept of domestication 18
carbon dioxide transfer rate (CTR) 117 conditioning for the market 10
carbon source 33, 172, 218 continuous cultivation 197
cardinal growth temperatures 35 continuous fermentation process 62
carotenoids 189, 199, 214 continuous process 220
catabolism 85 continuous stirred tank reactor (CSTR) 109, 234
catabolite repression 145 Contois kinetics 131
cell Convenience food 2
– density 165 cooling jackets 102
– dry mass 290 copper ions 239
– dry mass concentration 265 corn steep liquor 246
– membrane 60, 265 Corynebacterium 2
– pellets 289 Cosmetics 3
– walls 152 cost benefit ratio 319
Cell culture 31 co-substrates 154, 180
cell disruption 246 Crabtree effect 143
 Index | 337

cultivation parameters 287, 288 Ectoin 3

culture medium 41, 276 EDTA 52
cure diseases 29 electrical engine 274
cyanobacteria 183, 194, 195, 204 electrical potential 257
cytochrome 189 electrochemical
cytoplasm 183 – effects 254
– half- cells 257
D – reactions 258
dairy products 1, 21, 61, 148 – sensors 256, 259
Darcy’s law 166 electrostatic force (Coulomb force) 254
dark reaction 190 elemental
decay phase 87 – balance 42, 46
decomposition of cellulose and hemicellulose – composition 44, 58
231 – formula 59
degrees of freedom 306, 308 enantiospecific reactions 230
dehydrogenases 246 endogenous respiration 87, 137
density of the fluid 98 energy
desulfurization process 235 – supply 104
deterministic rules 298 energy and mass transfer 104
detoxification 242 energy crops 235
detoxification processes 240 energy input 104
diatoms 321 energy producing metabolic pathways 84
diauxic energy source 33
– growth 134, 153 energy transfer 107, 110
– lag phase 136, 142 environmental foot print 139
diffusion 93, 109, 110, 116, 117, 126 enzymatic
– coefficient kL 259 – activation 173
digestible cell wall 185 – cleavage of the residual 243
dilution rate 221, 310 – reactions 69
diploid 163 – system 20
disperse systems 21 enzymatic hydrolysis 58
dissociation constant 52 enzyme 72
diversity 22 – kinetics 130, 298
DNA 187 enzyme kinetics 83
domestication 219 enzymes 1, 6, 7, 11, 19, 68, 70, 231, 246, 269
downstream – kinetics 35
– costs 116 – technical see enzyme
– operation 141 epigenetic level 310, 315
– processing 10, 40, 152, 163, 174, 246, 247 equilibrium 51
– stage 12 Erlenmeyer flasks 278
drugs 4 essential amino acids 213
dynamic modelling 319 ethanol 7, 115, 122, 131, 134, 137, 142, 152, 154,
dynamic pressure method DPM 125 164, 226, 307
dynamic viscosity of the medium 209 – inhibition 140, 145, 227
– production 142
E eukaryotes 164, 175
E. coli 18, 73, 131, 145, 150, 176, 269, 292 eukaryotic
economic efficiency 94 – microalgae 190
ecosystems 220 – proteins 176
338 | Index

eutrophication of the environment 238 food

exhaust gas 285 – supplements 3
exoskeletons 242 food stabilization 148
exponential decay of light intensity 317 food supplement 61, 185, 211, 213, 215
exponential growth 160 force of viscosity 165
exponential growth phase 279, 287 Fourier’s law 102, 166
exponential phase 86 freshwater algae 195
expression system 31 functional foods 3
external cooling circuit 275 fungi 96, 106, 108, 174
extracellular
– technical enzymes 176
extracellular expression systems 174 G
extracellular product 109 G6P isomerase 134
extracellularly produced enzymes 229 gas flow rate 104
extreme milieus 23 gas mass flow measurement 253
extremophiles 240, 292 gas outlet filter 281
gas transfer 208
gas transfer coefficient kLa 120
F
gaseous substrate 238
fed-batch 177
gas-liquid mass transfer 95
– phase 276
genetic engineering 164, 184, 246, 293
fed-batch processes 157, 180, 275
genetic stability 61
feed controller 275
Gluconobacter oxydans 243
feed substrate concentration 158
glucose 138, 142, 172, 185, 190, 197, 278, 304
feeding rate 160
– and acetate depletion 177
feeding strategy 157, 161
fermentation 18, 117, 273 – feed 287
– microbial 7 – sensor 269
– process 2, 272 – uptake 122
– vessels 103, 283 Glutamate 2
fermentation process 140 glyceraldehyde-3-phosphate (G-3-P) 190
fermentative biogas 231 glycerol 197
Fermented foods 1 glycolysis 304, 308
Fibonacci sequence 76, 80 glycosylation 175, 176, 293
Fick’s law 166 Golgi apparatus 186
Fick’s law of diffusion 119 Gram staining 21
field-effect-transistor (FET) 269 graphical model 295
filamentous fungi 80 green gap 200
filter cake 166, 173 growth 66
filtration 168, 277 – and production phase 228
first generation biofuels 139 – equation 80
flagella 187 – experiment 77
flocculation 165 – factors 60
flow profile 99 – model 297
flow regime 104, 105 – rate 58, 67, 80, 84, 106, 122, 131, 178
fluorescence 266 – velocity 60
fluorescence markers 215 – yield 147
foaming 41, 54, 106, 282 Growth factors 56
focus effect 209 growth medium 32
 Index | 339

H inlet gas 273

Hagen–Poiseuille equation 209 in-line measurements 251
haploid cell 162 inoculation 225, 276, 287
harvesting 165 inputs and outputs of gas 272
health care 320 Insulin 22, 176
heat of combustion 216 integrated continuous bio-manufacturing
heat transfer 95, 102 process 247
heat-exchanger network 285 integration 174
heavy metal resorption 240 – sciences 13
Henry’s law 117 – society 2, 4
heterofermentative bacteria 148 μ-integration 318
heterotrophic 42 integration into society 320
– bioprocesses 182 intermig 100
– cultivations 195 intermittent light effect 319
heterotrophic growth 185 intracellular
heuristic models 297 – enzyme system 176
hexoses 135 – fluxes 308
high performance liquid chromatography (HPLC) – regulation 169
289 – stoichiometry 136, 142
high salt concentrations 186 – storage carbohydrate 163
homogeneous light distribution 197, 209 intracellular enzymatic steps 82
hyaluronic acid 3, 153 intracellular osmotic compounds 131
hydraulic retention time HRT 234 intrinsic security 216
hydrodynamic time constant 225 invertase 55, 57
hydrogen bonds 152 invertase activity 134
hydrogen producing bacteria 233 ion-sensitive electrode 255
hydrolases 20 irradiance 191
hydrolysates 164 isomers 245
hydrolysis of water 259
hydrolytic activities 55 K
hydrolytic bacterial species 232 kinetic equations 85
hydrolytic enzymes 174 kinetics 66, 111
hydrophobic 151
hydrophobic materials 258 L
hydrostatic pressure 208 lab work 8
hygienic design 94, 102 lactate 150, 153
lactic acid 7, 57
I Lactobacilli 61
immobilization 229 lactose 172
immobilized cell pellets 108 LacZ gene 269
impeller 93, 95, 99 lag phase 86, 116, 131, 136
incident light intensity 201 Lambert–Beer law 263, 300
inclusion bodies 176 laminar flow 101
incomplete oxidation 247 Liebig’s law of minimum 46
industrial applications 95 ligand exchange solvent extraction 239
information flow 13 light 191
infrared range (IR) 192, 254 – absorption 198, 315
inhibiting by-products 155 – distribution in the reactor 318
inhibitor 72, 74, 139 – energy collection 188
340 | Index

– harvesting complexes 188, 316 medium 41, 44, 48, 53, 58, 93
– inhibition 207 – conditions 154
– intensity 199 – design 111
– reaction 189 mesophilic 23, 234
light reaction 317 metabolic energy balance 298
Lignocellulose 58 metabolic modelling 82
limiting element 48 metabolic networks 302
limiting factor 46, 125 metabolic turnover rates 246
linear electron transport chain 189 metabolites 173
linear flux models 309 methane 62, 116, 235
linear growth model 82 methanol 61
linear independent balance equations 303 Michaelis–Menten
linear optimization 310 – Equation 75
linear programming 311 – kinetics 71, 131, 134
linoleic acid 213 Michaelis–Menten kinetics 74, 86, 199, 223,
lipid droplets 187 303
lipids 134, 213 microalgae 29, 114, 116, 182, 210, 211, 236,
lipophilic extraction 214 238, 264, 314, 317, 321
local irradiance 209 – species 184, 194
logistic growth curve 82 – strains 183
Lotka–Volterra models 301 microbial
Luedeking–Pirt equation 85 – applications 95
luminostat 203 – growth 33, 250, 302
Lysogeny broth (LB) 53, 54 – metabolism 37
lysosomes 187 microfluidic devices 111
micronutrients 47
M microorganisms 321
macroalgae 182 micro-sensors 111
macromolecular composition 58, 297, 319 microtiter plates 111, 260, 271
macromolecules 173, 195, 262 mini-harvest protocols 172
macronutrients 47 minimization of energy losses 141
magnetite nano-particles 240 mitochondria 183, 187, 314
maintenance 34, 84, 87, 198, 203, 227, 228, mitochondrial proton gradient 133
308 mixing 93, 95, 185
maltodextrin 140 model
mammalian cells 32, 114, 176, 215, 266 – building process 297
manometer 281 modeling
mass action law 51, 68, 70, 87 – of bioprocesses 295
mass transfer 131 molasses 54–56, 139, 164
material and energy flows 273 molecular sieve 141
material and energy fluxes 93 Monod
mathematical models 296 – equation 83, 130
maximum specific growth rate 179, 199 – kinetics 131, 199, 318
mean residence time 221 monomers 174
mechanical process engineering 165 moss 292
media 165 Moving Bed Biofilm Reactor (MBBR) 109
medical applications 319 mycelium 26, 170
 Index | 341

N oxygen transfer rate (OTR) 110, 120, 162, 227

Navier–Stokes’ equation for sedimentation oxygen uptake rate OUR 117, 142
velocity 165
Nernst equation 255 P
network structure 302 pant cell 145
nicotineamide adenine dinucleotide (NAD+) 53, parallel mini-reactors 112
73 partial pressure of CO2 197
Nitrate 51 Pasteur effect 142, 162, 293
nitrification 238 pellet 26
nitrogen Penicillin 4, 26, 110, 168, 243, 293
– balance 45 Penicillium 170
– limitation 216, 266, 317 performance molecules 151
– sources 43, 135, 164, 172, 238, 305 peristaltic pumps 282
nitrogen to carbon ratio (C/N ratio) 234 permeate flow 229
NMR technology 309 Pestoria 333
nodes of the network 302 petri dishes 321
nominal working volume 160 pH 38, 52, 82, 92, 154, 179, 203, 226, 255, 258,
nonlinear algebraic equations 221 268
non-photochemical quenching (NPQ) 199 – measurement 260
nucleic acid 43, 163 – sensor 256, 282
nucleus 163, 186 pH value 37
nutritional value 211 pharma industry 180, 247
pharma-proteins 110
Phosphate 50, 57, 136, 180, 278
O
– feeding profile 288
off-gas analysis 252
– limitation 314
off-gas measurement 124 – starvation 287
on-line sensors 251 phosphate substitute 151
open ponds 183, 204 phospholipids 213
optical density OD 264, 289 photobioreactors 29, 192, 203, 322
optical sensor 260 photo-bioreactors 183
optimization 297 photoconversion efficiency (PCE) 216
– approach 314 photon energy 193
– paradigm 315 photon flux density (PFD) 197
optimum conditions 80 photons 200, 206, 319
optimum harvesting point 87 photosynthesis 28, 134, 139, 186, 190, 218, 315
optimum working point 227 photosynthesis irradiance curve (PI-curve) 197
organic biopolymers 231 photosynthetic
organic loading rate OLR 234 – light reaction 188
O-ring 103, 274 – organisms 184
osmotic pressure 38, 41, 55, 145, 194 photosynthetically active radiation (PAR) 192
overpressure 286 phototroph organisms 95
– security valve 281, 284 phototrophic 116
overshoot metabolism 147, 150, 157, 161, 176, phototrophs 199
180, 246, see also Crabtree effect phylogenetic group 23
oxygen physiological behavior 314
– limited growth 137 phytase 272, 287
– turnover rate 143 Pichia pastoris 178
oxygen transfer 94 PID control 296
342 | Index

piezoelectric effects 253 PUFA 3, 212

pigment 197, 199, 214 pump calibration 253
pilot scale 273 pyrenoid 186
Pirt equation 223, 227, 303 pyruvate 304
Planck constant 193
plant cell 136 Q
plasmids 18 quantum yield 218
plastics 7
platform molecules 151 R
plug flow reactors 209 racemic mixtures 230
pneumatic reactors 104 raceway ponds 204
pO2 111, 142, 210, 258, 259 reaction rate 35
– electrode 269 reactor materials 208
– sensor 282, 287 recombinant proteins 30, 174, 179, 215
polylactides 7 rectification 141
polysaccharides (PS) 152, 183, 319 recycle ratio 230
population balance 301 redox
posttranslational modifications 216 – balance 307
power input 97, 99, 114, 204, 210 – equivalents 307
power number 97 redox balance 133
power-to-fuel 62, 142 redox reactions 237, 240
practical use 4 regioselectivity 243
preculture 278, 287 regulation 25
predator-prey systems 302 renewable
pressure sensors 253 – biomass 182
process – hydrocarbon feedstocks 238
– development 112 renewable biomass 29
– engineering 12, 13, 18 renewable resources 153, 320
– integration 141 repeated batch 154
– safety 18 respiration 52, 116, 304, 310
– stability 18 respiratory chain 134, 314
process analytical technology (PAT) 250 respiratory quotient RQ 122, 310
process control 261, 275 Reynolds number 97, 101
process control design 161 ribosomes 175, 187
process control system 283 ribulose-1,5-bisphosphat-carboxylase/-
process development 129 oxygenase (RuBisCO) 186,
process engineering 92, 163, 182, 218 196
process monitoring 253 ribulose-1,5-bisphosphate
product formation 66 carboxylase/oxygenase (RuBisCO) 190
product inhibition 131 Rushton turbine 96, 99, 100
product quality control 248
production costs 140, 186 S
production phase 180 S. cerevisiae 131
promotor 175, 178 scale-up 297
protein scattering 201, 263
– content 266 sea water algae 195
– content in microalgae 213 secondary metabolites 29
proteins 43, 152, 166, 174 secretion 179
protocol for sample acquisition 291 sedimentation velocity 165
 Index | 343

selectivity 20 stirrer 93, 94

self-regulation mechanism 234 stock solution 48
self-reproduction 14 stoichiometric
self-sterilization 24 – coefficients 44
sensitivity to shear stress 96 – data 44
sensors 250, 254, 255, 273 stoichiometric matrix 306, 310
shaking flasks 94, 112, 125, 250 stoichiometry 42, 58, 66, 228, 237, 253, 297
shear Stokes law 165
– forces 170 strain 22, 25, 32, 41, 111
– sensitive cells 107 – robustness 292
– sensitive microorganisms 99 – selection 18
– sensitivity 185 structure 33
– stress 100, 125, 177 struvite 52
shear rate 99 submers cultivation 108, 150, 169
siderophores 52 submersed membranes 229
signals 15 substrate
simulation 295 – balance equation 159
single-use bioreactors (SUB) 113 – feeding stream 157
slug flow 105 – inhibition 75, 131, 180
solar fuels 214, 216 – limited growth 137
solids retention time SRT 234 – specificity 68
solvent recycling 245 – turnover rate 71
sorbitol 245 – uptake 85, 129, 137
sparger 95, 104, 117, 274, 280 – uptake kinetics 164
spatial order 15 – uptake limitation 146
spatial self-organization 241 – uptake rate 84
specific cake resistance 166 substrate limited growth 83
specific costs 111 substrate uptake rate 222
specific glucose uptake rate 142 substrates 93, 179
specific growth rate 79, 86, 87, 134, 135, 162, surfactants 4
171, 222, 310 symbiogenesis 187
specific substrate uptake rate 199 syngas 238
specific uptake rate 83
spectrometer 263 T
spectroscopy 262 taxonomic group 183
Spirulina 184, 203, 211 technical applications 7
ß-carotene 186 technical maintenance 180
starch 139, 153, 216, 267 technical media see medium
– granules 187 technical transport/reactions system 92
starter cultures 61, 149 temperature 275
stationary phase 86 – sensitivity 258
steady state 70, 124, 220, 225 – sensor 283
stems 183 temperature dependent 35
stereospecifity 230 tensides 4
sterile filter 284 theoretical working point 224
sterilization 251, 276, 284 thermodynamics of the system 129
sterilization in place (SIP) 94 thermophilic 234
stillage 141 Thiobacillus ferrooxidans 239
stirred tank reactor (STR) 94, 104, 273 thylakoid 187
344 | Index

time-continuous model 300 volumetric

total flow out 229 – biogas productivity 236
toxic effects 131 – concentrations 66
toxins 184, 211 – gas flow rate 98
trace elements 55, 151, 277 – heat flux 102
transducer 268 – oxygen demand 125
transitional system dynamics 225 – productivity 66, 154, 223
transport and conversion steps 39 – reactions rates 68
transport level parameters 110 volumetric flow rate 158, 282
transport/reaction system 199, 224 volumetric gas flow 117
triacylglycerol (TAG) 214, 216
tubular reactors 208 W
turbidostat 203 wash-out case 222
turnover rates 66, 132 wastewater 33, 269
two film model 120 – treatment 109, 110, 114, 238
wavelength 191
U whey 148
ultraviolet range (UV) 192 wine making 134, 149
unbudded cells 166
undirected mutagenesis 169 X
unicellular plants 27 Xanthomonas 2
Upflow Anaerobic Sludge Blanket reactor (UASB) Xanthomonas campestris 153
109
upstream processing 276 Y
upstream stage 11 yeast 18, 38, 55, 61, 80, 96, 122, 140, 162, 174,
178
V – catabolism 310
vertical shaped bioreactor 95 – growth 313
Vinegar 2 yield 66, 84, 85, 87, 138, 155, 159, 288
Vitamin B12 53, 56, 195, 247 – parameter 132
Vitamin C 3, 57, 243
volatile fatty acid (VFA) 233 Z
volatile organic compounds 117 Zymomonas mobilis 226