You are on page 1of 156

Cultivation Technology

Gen Larsson

Dept. Industrial Biotechnology, School of Engineering Sciences in Chemistry, Biotechnology and Health,
Royal Institute of Technology (KTH), Stockholm, Sweden
The simulations are all performed in “SimuPlot”, a Matlab based software developed for
bioprocess calculation and optimisation by Professor Emeritus Sven Olof Enfors, KTH.
His personal webpage (enfors.eu) has more information about SimuPlot and Fermentation
process technology.

10th edition, Stockholm 2020


Contents

Contents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iii

1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1 Definitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 The historical eras of biotechnology . . . . . . . . . . . . . . . . . . . . . . 2
The first biotechnology era . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
The second biotechnology era . . . . . . . . . . . . . . . . . . . . . . . . . 5
The third biotechnology era . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
The fourth biotechnology era . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.3 The biotechnology industry . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.4 Price/volume relations of biotechnology products . . . . . . . . . . . . . . 9
1.5 Product categories . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.6 A comparison between bioproduction and organic synthesis . . . . . . . . 11
1.7 The industrial bioprocess . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12

2 Production Cells . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.1 Eukarya . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.2 Archaea . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
2.3 Bacteria . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
Gram negative bacteria . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
Gram positive bacteria . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
Cyanobacteria . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20

3 Cultivation Media . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
3.1 The metabolic starting point of medium design . . . . . . . . . . . . . . . 24
3.2 Medium categories . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
3.3 Medium design based on the elementary composition of a cell . . . . . . . 27
3.4 Substrates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
Substrates including macro elements . . . . . . . . . . . . . . . . . . . . . 28
Substrates including the micro- and trace-elements . . . . . . . . . . . . . 31
Water . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
Control of culture pH . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
Calculation of a simple defined medium . . . . . . . . . . . . . . . . . . . 33
Sterilisation of media and medium components . . . . . . . . . . . . . . . 34

4 Growth of Cells . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
4.1 Growth forms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
4.2 Storage of production cells . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
4.3 Inoculum preparation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
4.4 Growth patterns . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
The phases of a growth curve . . . . . . . . . . . . . . . . . . . . . . . . . . 42
4.5 The rate of growth . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
4.6 Influence on the growth curve of dead and/or non-viable cells . . . . . . . 46
4.7 Factors affecting the maximum growth rate . . . . . . . . . . . . . . . . . . 47
4.8 The time-dependent uptake of sugars . . . . . . . . . . . . . . . . . . . . . 49
4.9 The substrate uptake rate as a function of substrate concentration . . . . . 51
4.10 The Monod equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
4.11 Batch growth with limiting substrates other than the carbon source . . . . 55
4.12 Growth on mixed carbon or mixed nitrogen substrates . . . . . . . . . . . 57
4.13 Cellular production of inhibitory substances - overflow metabolism . . . . 58
4.14 Techniques to measure growth of cells . . . . . . . . . . . . . . . . . . . . 60
Direct counting of cells . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
Light scattering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
Weighing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
4.15 Measurements of medium and intracellular components . . . . . . . . . . 63

5 Yield and Productivity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65


5.1 Yield . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
5.2 Deviations from a constant yield . . . . . . . . . . . . . . . . . . . . . . . . 69
Growth at very low glucose concentrations . . . . . . . . . . . . . . . . . . 69
The yield coefficient at high glucose concentration . . . . . . . . . . . . . . 73
5.3 Productivity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
5.4 The specific production rate . . . . . . . . . . . . . . . . . . . . . . . . . . 74

6 Bioreactors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
6.1 The stirred tank reactor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
6.2 Gas-mixed bioreactors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
6.3 Other types of bioreactors . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
Packed beds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
Wave bioreactors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
6.4 Measurement and control of bioreactors . . . . . . . . . . . . . . . . . . . 83
In situ analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
Liquid sampling for external analysis . . . . . . . . . . . . . . . . . . . . . 85
Gas sampling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
Supervising the bioreactor performance . . . . . . . . . . . . . . . . . . . . 85

7 Agitation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
7.1 The use of balance equations . . . . . . . . . . . . . . . . . . . . . . . . . . 87
7.2 Macroscopic momentum balance describing the macromixing . . . . . . . 89
7.3 Molecular transport balance describing the macromixing . . . . . . . . . . 90
7.4 Mixing devices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
Agitator types . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
7.5 Turbulence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
7.6 Measurements of mixing . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98

8 Cultivation Concepts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103


8.1 The general mass balance concept . . . . . . . . . . . . . . . . . . . . . . . 103
The balance of substrate over the reactor . . . . . . . . . . . . . . . . . . . 105
The balance of cells and product over the bioreactor . . . . . . . . . . . . . 105
8.2 Batch processing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
8.3 Continuous cultivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
Continuous cultivation in practice . . . . . . . . . . . . . . . . . . . . . . . 108
The general liquid side mass balance for a chemostat . . . . . . . . . . . . 109
The balance of cell mass in a chemostat . . . . . . . . . . . . . . . . . . . . 109
Mass balance of the limiting substrate . . . . . . . . . . . . . . . . . . . . . 110
The mass balance of other substrates than the limiting . . . . . . . . . . . 111
A plot of parameters against the dilution rate . . . . . . . . . . . . . . . . . 112
Reasons for restriction of the growth rate . . . . . . . . . . . . . . . . . . . 114
The chemostat performance . . . . . . . . . . . . . . . . . . . . . . . . . . 114
Effects of the saturation constant . . . . . . . . . . . . . . . . . . . . . . . . 115
The self-regulating metabolic control of the chemostat . . . . . . . . . . . 116
Applications of the chemostat . . . . . . . . . . . . . . . . . . . . . . . . . 117
Continuous cultivation with recirculation . . . . . . . . . . . . . . . . . . . 119
8.4 Fedbatch cultivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
Mass balances in fedbatch cultivation . . . . . . . . . . . . . . . . . . . . . 121
The balance for cell mass and product in fedbatch cultivation . . . . . . . 121
The mass balance for substrate in fedbatch cultivation . . . . . . . . . . . . 122
The substrate feed . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122
The maximum cell density . . . . . . . . . . . . . . . . . . . . . . . . . . . 127

9 Aeration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
9.1 Parameters controlling the oxygenation . . . . . . . . . . . . . . . . . . . . 129
9.2 The macroscopic mass balance of oxygen . . . . . . . . . . . . . . . . . . . 133
9.3 The molecular oxygen transfer rate, OTR . . . . . . . . . . . . . . . . . . . 134
9.4 Spargers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
9.5 Methods to estimate the kL a . . . . . . . . . . . . . . . . . . . . . . . . . . 142
The gas balance method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142
Gassing out with nitrogen . . . . . . . . . . . . . . . . . . . . . . . . . . . 144
The dynamic method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145
9.6 Mixing and aeration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147
Introduction 1
1.1 Definitions . . . . . . . . . . . . 1
1.1 Definitions 1.2 The historical eras of biotechnol-
ogy . . . . . . . . . . . . . . . . . . . 2
The scientific field that we call “Biotechnology” sometimes refers to a The first biotechnology era . . 3
wide variety of different issues. In this book we will use the following The second biotechnology era 5
meaning of this word: The third biotechnology era . 6
The fourth biotechnology era 7
Biotechnology is the integration of natural sciences and engineering in order to 1.3 The biotechnology industry . 8
achieve the application of organisms, cells, parts thereof and molecular analogues 1.4 Price/volume relations of
for products and services (European Federation of Biotechnology, 1980). Further, biotechnology products . . . . . . 9
the United Nations Convention on Biological Diversity defines Biotechnology as: 1.5 Product categories . . . . . . 10
"Any technological application that uses biological systems, living organisms, or 1.6 A comparison between biopro-
duction and organic synthesis . . 11
derivatives thereof, to make or modify products or processes for specific use."
1.7 The industrial bioprocess . . 12
Clearly, the subject involves the use of cells or their various building
blocks in order to make things that are valuable to society.
The word “Biotechnology” is a combination of “Bio” and “Technology”
and it is clear that in order to be truly successful in this field a person
needs to be skilled not only in the field of biology and its subdivisions
but also in engineering-related subjects. The Bioprocess Engineer works
with the design of processes which build on the reaction mechanisms of
cells or enzymes. The goal is to arrive at economically viable and safe
processes where the products meet the desired quality criteria. Whereas
process safety and economy are relatively "easy" to define, quality is
always specific for a given product and its applications. In this book,
we shall consider bioprocesses that are based on living cells. Enzymatic
catalysis outside the context of a whole cell is therefore beyond the scope
of this book.
The word “Fermentation” is by definition a process where an important
end product is accumulated by a microorganism and where an organic
molecule of the primary metabolism is used as terminal electron-acceptor
instead of an external electron acceptor (oxygen, nitrate etc.). Fermenta-
tion is therefore an anaerobic process. Examples of such products are
acids and solvents such as: lactic acid, ethanol, propionic acid, acetone,
butanol and methane. Historically, it was these fermentation products
that constituted the starting point for Biotechnology development more
than 10 000 years ago. Consequently, these products were named “fer-
mentation” products, the bioreactors were called “fermenters” and the
area became known as “Fermentation Technology” (Swedish: “jäsningspro-
dukter/fermentorer/jäsningslära”). In this book, we strictly separate the use
of "cultivation" for growing and using microorganisms in bioreactors
and "fermentation" for metabolism of for instance sugars in the absence
of external electron acceptors. However, in the biotechnology field the
term fermentation is for historical reasons often also used synonymous
with cultivation.
2 1 Introduction

1.2 The historical eras of biotechnology

Biotechnology is one of the oldest technological fields known to man. This


is due to the fact that one of the application areas is the production and
preservation of food and beverages, to satisfy the primary needs of human
beings; things we cannot live without. Simultaneously, biotechnology
rapidly developed due to the search for methods and products to cure
illness.
The first discoveries that relate to Biotechnology originate from archae-
ological findings in the Middle East. The first people moving to the
Middle East were part of a migration of animals and humans caused by
climate change. The ice over Europe was melting and the climate was
starting to be suitable for survival. In the “Fertile crescent” (Figure 1.1)
12000-year-old traces of the life of our ancestors have been found. Other
such early settlements were successively formed around the globe where
Figure 1.1: Magnification of the geographical
area of the “Fertile Crescent” which is gener-
the ice from the last Ice Age had withdrawn and where conditions were
ally described by today’s Israel and Palestine equally favourable for settlements, such as e.g. southern and northern
territories, parts of southern Turkey, Kuwait, China, sub-Saharan Africa, Southeast Asia, Peru and Mexico (Figure
Lebanon, Jordan, Syria and most of Iraq. When
early settlement took place this was largely the 1.2).
area inhabited by the Sumerians. The country
between the rivers Euphrat and Tigris was Farming communities arose in the Levant and spread to Asia Minor,
called Mesopotamia between 6000-1500 BC. north Africa and the north of the Fertile crescent in 9500-9000 BC. The
first cities arose around 8500 BC. During these early times, the area
benefited from the rainfall due to the large mountain ridges in the north
and east, from the Taurus Mountains in Turkey to the Zagros Mountains
in southern Iran, and it was on these slopes that the first settlements
took place. The Sumer region was regularly flooded by the Euphrates
and Tigris rivers during these times. Knowledge and techniques on how
to effectively handle this natural resource was developed in the 5th
century BC. The irrigation method eventually adopted was the same
as the techniques used in Egypt in the area of the Nile. This included
Figure 1.2: Spreading of settlements from the
the digging of channels and arranging of reservoirs so that fields could
area of the “Fertile Crescent”(in red) which be flooded and drained upon demand allowing intensive large-scale
is named “the Cradle of Civilisation”. James cultivation of land.
Henry Breasted, Univ. of Chicago, seems to
have been the first to use the term in 1906 in his Apart from the climate and the rich wild life, another large benefit was
“Ancient Records of Egypt”. Numbers refer to
years BC. the occurrence of wild crops. The Levant was during these times not dry
and barren but a fertile area. Early Neolithic farming was limited to a
narrow range of plants and the first cereal crops used were different types
of wheat such as einkorn and spelt, which grew wild on the slopes of the
mountains. Also barley grew here, which had the advantage that it was
rich in amylase (which will rapidly degrade the starch), could withstand
harsh climate changes and had a short growth period. Early on, farming
was restricted to sheep and goats and by 8000 BC this included also the
domestication of cows and pigs. These developments made food available
year round, which allowed for a more dense population. The techniques
for e.g. grinding of seed, ploughing and the storage and preservation of
food rapidly followed. The latter involves several biotechnology-based
techniques, as we shall see.
1.2 The historical eras of biotechnology 3

The first biotechnology era

The first Biotechnology era we could consequently call “the Ancient Era”
and it is very probable that one of the first biotechnical products was
bread. This bread was however nothing like the bread that we use today
when the use of baker’s yeast leads to raising of a dough which, in turn,
leads to a product which is soft and a bit “spongy”. The ancient bread
on the other hand was hard to chew and most likely also sour since Figure 1.3: Clay tablet showing the first
lactic acid bacteria, that are present everywhere in nature, could not steps in making beer from bread. Egypt
1300 BC.
be avoided in the baking process. The intentional use of leaven (sour
dough) to raise dough has continued throughout the years and has for
many centuries been coupled to a healthy life style. The health issue was
actually so pronounced that in Germany it was forbidden to use baker’s
yeast for bread-making until AD 1600 when it was allowed but under
severe criticism mainly from physicians. The German “Purity law” (the
“Purity order” implemented in 1516 and updated in 1993) still applies
for the production of beer where only barley-malt, water, hops and yeast
are allowed in the brewing, although today maybe not only for health
reasons.
Abbreviated text extract from “A Hymn to
The making of beer is also one of the oldest biotechnological processes Ninkasi”, 1800 BC. Black JA, Cunningham G,
known. During the early days of man, the knowledge of how to make beer Fluckiger-Hawker E, Robson E and Zólyomi
was considered a sign of a highly civilised person. Bread was frequently G. The Electronic Text Corpus of Sumerian
Literature, http:// www-etcsl.orient.ox.ac.uk.
used as the raw material for beer making and was crumbled and dissolved
Oxford 1998
in water to form a loose porridge. This was filtered and from this liquid,
1-4. Given birth by the flowing water (. . . ),
a wild fermentation based on the everywhere present yeast took place.
Tenderly cared for by Ninhursaja,
The word “bappir” was used for this beer/bread in Mesopotamia. The 5-8. Having founded your town upon vax,
same procedure is still used in for instance Egypt ("Bowsa") and Ethiopia She completed its great walls for you,
("Sowa"). 9-12. Your father is Enki, Lord Nudimmud,
Your mother is Ninti, the queen of the
Archaeological excavations have found many artefacts which are claimed abzu,
to be reminiscent of beer and beer-making (Figure 1.3). The discoveries 13-16. It is you who handle . . . . . . and
dough with a big shovel, mixing, in a pit,
have been either direct, i.e. vessels containing liquids with traces of grain the beerbread with sweet aromatics,
and oxalate, or indirect, e.g. clay tablets describing the beer-making 17-20. It is you who bake the beerbread in
procedure. Whether we would call these liquids “beer” is a matter of the big oven, and put in order the piles of
hulled grain,
our definition of this product. Early traces of fermented beverages can
21-24. It is you who water the earth-
also be found in China where a drink made of rice, honey and fruit was covered malt; The noble dogs guard it
produced as early as 7000 BC. The earliest traces of beer, on the other even the potentates,
hand, we find in countries which produced the necessary grains for this 25-28. It is you who soak the malt in a jar
The waves rise, the waves fall.
particular product. The technique of making fermented products was
29-32. It is you who spread the cocked
most likely spread by the migration of people between countries, but we mash on the large reed mats; coolness
also know that there was a considerable export taking place whereby e.g. overcomes.
the Egyptian could consume Babylonian beer. 33-36. It is you who hold with both hands
the great sweetwort, Brewing it with
The beer-making process is described on a very famous clay tablet which honey and wine
37-40. /damaged line/ You . . . . . . . . . the
bears a “Hymn for Ninkasi” – the Sumerian Goddess of Beer and chief
sweetwort to the vessel.
brewer, which is dated 1800 BC. The beer-making process, which is 41-44. You place the fermenting vat, which
described into some detail, includes all steps of the modern process. makes a pleasant sound, appropriately on
The text of the clay tablet is shown in the margin. Try to identify the top of a large collector vat.
45-48. It is you who pour out the
beer-making steps!
filtered beer of the collector vat; It is like
During these early times, the religious part of beer-making also comes the onrush of the Tigris and the Euphrates.

through very strongly and the promotion of a “Goddess of Beer” suggests


the level of seriousness this issue had. The state of drunkenness was
believed to result also in visionary abilities and could lead to telling of
4 1 Introduction

prophecy, which was considered a most valuable quality of beer and


is very much dwelt upon in Sumerian literature. The religious and the
healing elements together with the quest for food were therefore very
strong driving forces for technical development at this time.
Extract from Lungalbanda and the Anzu
Bird (Gilgamesh Epic), 2100 BC. Lugalbanda As mentioned in the very first paragraph of this book, the quality of a
was the father of Gilgamesh. The Electronic product is very important. The quality of the ancient beer would today
Text Corpus of Sumerian Literature, Oxford probably be a subject of discussion and it is very likely that we would
University. http://etcsl.orinst.ox.ac.uk/cgi-
not find it tasty. One reason is that it was difficult to remove all the grain
bin/etcsl.cgi
from the beer by the then available filtration methods - a lot of husks
Her fermenting-vat is of green lapis lazuli,
were probably still floating at the surface. In this “cake”, microorganisms
Her beer cask is of refined silver and gold.
If she stands by the beer, there is joy, would easily grow since all ingredients are there: the oxygen from the air,
If she sits by the beer, there is gladness; which also carries spores of mold, and the carbohydrates left from the
As a cup-bearer (Swedish: munskänk) she grains - central elements for the propagation of mold. This is supported
mixes the beer,
by for instance figures on grave seals that show the use of long straws to
Never wearying as she walks back and
forth, sip beer, probably in order to penetrate and not disturb the cake at the
Ninkasi, the keg at her side, on her hips; top (Figure 1.4).
May she make my beer-serving perfect.
When the bird has drunk the beer and is A further taste issue was that lactic acid bacteria probably contaminated
happy, the beer, especially if it was based on breadcrumbs. This probably made
When Anzu has drunk the beer and is
the beer much more sour than the beer produced today. The quality was
happy,
He can help me find the place to which therefore hampered by a lack of knowledge of the presence of microor-
the troops of Unug are going, ganisms. The fact that microorganisms were not known also means that
Anzu can put me on the track of my pure cultures could not be used nor controlled. Contaminations from the
brothers
environment likely always took place.
Due to the frequent occurrence of lactic acid bacteria, sour milk belongs
also to the first biotechnological products. Cows and goats gave milk that
could be used for food directly, but contamination with these bacteria
was likely a frequent occurrence. In the worst case the contamination
could be with harmful microorganisms, but it is a good guess that the
sight and smell of the sour milk became the sign that the product was
healthy. Harmful microorganisms would not have made the milk sour
Figure 1.4: Beer serving and drinking in but might show signs of putrefaction (another biotechnology process!)
Sumer approx. 2000 BC. The enthroned female
is most likely served by the man in the middle that would have made it highly undesirable. The acidification meant also
and is sharing the drink with the man to the left. a drop in pH which helped in the preservation of the milk quality since
He is thought to be Gilgamesh who was the son
many detrimental microorganisms cannot grow under acidic conditions.
of Lugalbanda (see Gilgamesh epic) since the
scene takes place in an mountainous area, well Several inscriptions talk about the use of sour milk not only for drinking
known for the struggles of this character. Note but also for adding of flavour. A recipe from Mesopotamia dated about
the straws (to protect from taking the filth from
the top)!, which at this time (bronze age) were
3570 BC talks about meat spiced with onion, garlic, fat and sour milk.
probably made of metal. Many of the items in
the figure represents the contemporary gods:
The production of cheese is another early biotechnology advancement,
Innana/Ishtar, the Goddess of Love and War but it should be noted that this is basically an enzymatic process. Even if
who was associated with the city of Uruk (the it is likely not true, it is nice to think that the first cheese was produced
eight-point star, Venus); Enki (or Ea): the God
of wisdom who had given mankind the arts
from the storage of milk in the stomach of a calf which was used as a
and sciences as well as industries and manners carrier-bag in those days. In the stomachs of all mammals, the enzyme
of civilisation and was associated with the city chymosin and a mixture of different protein-degradation enzymes, are
of Eridu (the fish); Utu the sun God, who was
associated with the city of Sippar (circle); and available. The chymosine will rapidly hydrolyse the milk protein (i.e.
Nanna the Moon God who was associated with casein from the Latin “caseous” meaning “cheesy”) which separates the
the city of Ur (crescent moon). Coiled snake milk into a mass of fresh cheese (curd) and whey. However, it is believed
decorations at the side
that this production took place even before people started to domesticate
and breed animals.
Also, the production of wine is an early biotechnology process, but
it is believed to have developed comparatively later due to the fact
that wine-making requires breeding of the vine. The making of beer,
1.2 The historical eras of biotechnology 5

wine and also sake (the latter in some Asian countries) shows many
processing similarities, and it is easy therefore to see that they might share
a simultaneous development in time but at different locations around
the earth. All of them make use of the local production of raw materials
(barley, grapes and rice, respectively) which are rich in carbohydrates
and where appropriate microorganisms present in the vicinity can easily
grow. These will eventually all lead to the alcoholic fermentation that
takes place in the absence of oxygen.
Another example is the production of soy sauce in Asia. The production
is described in 2800 BC in China and it is still produced by the utilization
of Aspergillus oryzae, lactic acid bacteria and yeast growing on rice, which
in these countries is an abundant raw material.
In summary, many food processing processes were thus very early
established. It is very clear that the development of new products in
those early days, what we call “the First Biotechnology Era” was driven
by very powerful forces i.e. survival, religious beliefs, treatment of
illness and, most likely, also the search for substances to give pleasure.
The making of some of the products, e.g. beer and wine, has also
always had an ideological significance confirming national, cultural and
ideological affiliation. The documentation present in literature mostly
shows techniques and economic facts, but would consequently also
display a cultural pride.
Figure 1.5: Vine: here Grüner Veltliner, grapes
common in Austria and the Czech republic

The second biotechnology era

For the next generation of biotechnology-based products, we have to


wait a very long time i.e. to the middle and end of the 19th century,
which we might call “the Organic acid and Solvent era” and which
continued until just after World War II. Many products that we use today
were originally extracted from a natural resource and were at those
times not produced by living cells or by enzymatic transformation (and
were thus not “true” bioproducts”). One such example is the organic
acid; citric acid. This acid was originally extracted from lemons and
constituted a major industry in Italy starting in the 1890s. When it was
found that the mold, Aspergillus niger, could accumulate citric acid in the
surrounding medium, this replaced the extraction process completely
and the Pfizer company could start a commercial production based on
this new technique. The wide application of citric acid is based on the
fact that a 1% solution of citric acid has a pH of 2.2 and it is used in very
many industrial applications such as the preservation of food and for its
capacity as a complexing agent in metal treatment.
Another example from this era is the production of the amino acid
glutamate, which was initially extracted from wheat gluten or soy protein.
The production volume is today close to 1 000 000 tons per year and
glutamic acid is used as a much appreciated food flavour in Asiatic
countries (the fifth taste i.e. umami). The technological developments
based on the microorganism Clostridium glutamicum took place in the
1900s, but the process was only commercialised by the large Japanese
company Ajinomoto after World War II.
Figure 1.6: The microscope invented, drawn
and described by Robert Hooke in “Micro-
graphia”, 1664
6 1 Introduction

Why did we have to wait more than 8000 years between the first and the second Biotechnology era? The
answer is that we first had to see and understand the nature of microorganisms! The first discovery promoting
new developments were thus the high-resolution microscopes by Antonie van Leeuwenhoek (1632-1723)
and Robert Hooke (1635-1703), which were developed approximately at the same time. Leeuwenhoek’s
microscopes could magnify up to 200 times, which was extremely high for the time, and due to his skill in
making very good lenses. He was however very careful not to reveal this skill for competitive reasons and
this might have led to the slower developments despite the fact that the technology was already present.
What is so remarkable about this invention? At this time obviously nothing about growth and the origin of
microorganisms was known. The growth of comparatively larger organisms, which could be seen by the
naked eye such as worms etc, were also thought to develop by spontaneous generation – mice developed
from old rags, maggots from meat etc. In 1665, Robert Hooke published the book “Micrographia” and in
1684, Antonie van Leeuwenhoek published “Microscopical observations about animals on the scurf of teeth”
in Philosophical Transactions of the Royal Society of London. These events are considered to be the birth of
the subject of “Microbiology”. The ability to suddenly “see” the microflora provided the foundation for the
coupling between cause and effect with respect to microorganism and product. The early examples studied
were the coupling of yeast and alcohol fermentation and lactic acid bacteria and lactic acid production by
pioneers such as Louis Pasteur (1822-1895). During the years to follow it was also established that specific
microorganisms were responsible for the putrefaction and decay of food (J. Leibig 1839). In parallel, great
discoveries were being made in medicine with respect to the microbial origin of deadly diseases, and this
added to the foundation for development. Now, microorganisms could be observed and their features studied
and documented which marks one of the cornerstones of this era. The driving forces for development
during this period were thus the technical inventions that allowed the discovery, but also the isolation, of
microorganisms that could produce desired products. This provided the foundation for the development of
commercial processes based on microorganisms.
The commercial processes of solvents relate to World War I, when acetone was in great demand for use in the
production of cordite for gun powder. Acetone and the solvent, 1-butanol (used for automobile paints and as
a building block for esters) were previously produced exclusively from petrochemical raw material. This
raw material was however difficult to obtain during wartime. In the beginning of 1915, Chaim Weizmann,
who became the first president of Israel, developed a process for the production of acetone/butanol by the
microorganism Clostridium acetobutylicum. This process, the ABE (acetone/butanol/ethanol) process, was in
operation until the 1950s when oil became so cheap that the microbial process was no longer economic. This
shows that except for the necessary technical developments, war and the shortage of raw material as well as
economic factors are powerful driving forces for the development of alternative processes.
How were these producing microorganisms found? Common for them all is that they operate in specific niches
in nature and that they all make notable changes to the micro-environment in a way that is sometimes very
visible to the naked eye. The production of acids or solvents also means that no other microorganisms grow,
since the products are microbial inhibitors and a pure culture of the production microorganism naturally
develops. This is also the reason why large-scale industrial processing, satisfying the market demand for
these products, could be carried out although important techniques for the large-scale sterilization of air and
media were not yet available. The microorganisms and their products “sterilised” the environment.

The third biotechnology era

The third era is the “Era of antibiotics”, which are cellular products of the “secondary metabolism” as
compared to the solvents of the second era which are primary metabolites. The discovery and characterisation
of penicillin, which marked the start of this era, is attributed to Alexander Fleming in 1928. However, several
documented observations of the effects of penicillin were made already in the end of the 19th century. Howard
Flory and Ernst Boris Chain developed penicillin for medical use with the help of Norman Heatley (penicillin
analysis) that lead to production of the quantities needed for prolonged afficacy studies and treatment.
1.2 The historical eras of biotechnology 7

Flory and Chain shared the Nobel prize with Fleming for this discovery in
1945. However, the initial amounts of penicillin available at this time were
so small that they were not enough to fully treat patients. Limitations
that needed to be overcome for a large scale process included: (i) low
antibiotic producing strains, (ii) media that did not promote production
due to lack of essential side chains, (iii) penicillin is very heat labile,
which made purification difficult.
The first developments in penicillin production were based on strain im-
provements through random mutagenesis. The typical antibiotics process
requires rich media and large quantities of air. With process technology Figure 1.7: A. Characteristic “plaque” forma-
at the time it was difficult to maintain aseptic conditions, despite the tion around an antibiotic-treated pellet. No
cells will grow in the immediate vicinity. B.
inherent antibiotic properties of the product. Maintaining a pure culture Conidiophores with conidia (spores)
was very critical, since many strains of bacteria infecting the process
have the ability to degrade 𝛽 -lactame-based antibiotics (production of
𝛽 -lactamase). Methods for the sterilization of the cultivation medium and
for supplementation of sterile and clean air were therefore developed. The
production of penicillium is further based on the propagation of mold
and the formation of mycelium severely increases the viscosity of the
broth, which decreases mixing and oxygen transfer. The solution to these
problems heavily relied on engineering and constituted the cornerstones
of the new commercial processes and cultivation technology.

The fourth biotechnology era

Some species of microbial cells, as we have seen, naturally accumulate


products of interest and are able to secrete very large amounts of these
to the environment, provided the surrounding medium includes the
prerequisites for both growth and production. However, even if wild-type
cells, when they are discovered, are able to produce certain products, the
concentrations may not be large enough to allow a development of an
economically viable process.
The types of products above, such as penicillin and organic acids, refer
to production of small molecules but bioprocesses also include products
such as polymers like proteins. Enzymes, for example, are naturally
produced and excreted by Bacillus and different molds. This gives them
a competitive growth advantage in nature since it allows the hydrolytic
release of their own “food” through degradation of polymeric substances
in their natural habitats. To satisfy the ever-increasing demands, today’s
industrial processes have proceeded from the use of native strains to the
use of strains that have continuously been developed in the laboratory
over many years. Initially this work relied on a random treatment of cells
with e.g. chemicals and UV light in order to increase the rate of mutation
and in this way to find strains that had gained superior properties. The
new technical developments during the 4th era however opened the
possibility to introduce directed and rational mutations in cells. This
relied on developments in macromolecular science including the solving
of the structure of the
8 1 Introduction

nucleotides; DNA and RNA, the development of genetic manipulation techniques, as well as the understanding
of cellular control of protein synthesis. This provided the necessary tools for the strategic redesign of cells
with improved metabolism and for learning to overproduce recombinant proteins of different origin in
different microorganisms, what we call “recombinant DNA technology”. One of the early products from
this era is the recombinant human growth hormone, which is used as a pharmaceutical for treatment of e.g.
dwarfism.

1.3 The biotechnology industry

Biotechnology is used in a great number of different world-wide industrial sectors and its use is ever
increasing. By number of companies, the largest biotechnology sector in Sweden is the pharmaceutical sector.
Most of the companies are however very small, and most are in early product development phases and are
thus financed by investment capital. There are however few companies in Sweden producing pharmaceuticals
in particular with regard to biopharmaceutical drugs (protein drugs). Figure 1.8 shows the distribution of
companies in the so-called “new Biotechnology” or “Life-science” sector in Sweden (in 2011). This implies
that the large areas of traditional food production including the production of Baker´s yeast, wastewater
treatment and the production of biofuels are excluded.

Figure 1.8: The Life science area in Sweden (excluding food, biofuel production and wastewater treatment) and the distribution over different sectors.
Ref.: Life Science companies in Sweden. VINNOVA analysis; VA 2011:03. The area of “Bioproduction”, the third largest area in the figure, refers to a
small number of Swedish contract manufacturers (CMO´s or Contract Manufacturing Organisations) that are almost exclusively involved in GMP
production of biopharmaceuticals (proteins)

The production of chemicals and fuels by biotechnology is still comparatively low with respect to the number
of companies although the production volumes are large. This market is however rapidly increasing and
many companies involved in oil and polymer science are also looking at biotechnology-based processes.
The production of biofuels in Sweden today is largely directed towards ethanol and methane. Methane
production is by tradition taking place in connection to wastewater treatment plants, but also production
from solid and liquid industrial waste and manure is gaining a lot of interest. A grid of natural gas pipelines
is installed in southern Sweden and to this also methane gas from anaerobic digestion of waste biomass has
been connected after upgrading (removal of carbon dioxide).
1.4 Price/volume relations of biotechnology products 9

1.4 Price/volume relations of biotechnology


products
Table 1.1: Examples of biotechnology-
The biotechnology-based products today are indeed very many and the
produced products, their annual volume and
number is continuously increasing. In this section we shall try to group market price. (Ref.: Schmid R. 2000. Pocket
the products into categories to give an overview and to improve the Guide to Biotechnology and Genetic Engineer-
ing)
understanding of this interesting field. We shall start by looking at some
economic issues that all biotech products share. Production volumes Products Volume(t/y) Price(€/kg)
Beer 130 000 000 2.5
of biotechnology products vary greatly Table 1.1. At one end there are Ethanol 19 000 000 0.25

the multi-ton products, such as beer, and at the other, products that are Glutamic acid
Citric acid
800 000
700 000
1
1
required only in kilogram quantities annually, such as the pharmaceutical Detergent Protease 100 000 3
Aspartame 10 000 5
erythropoietin (EPO). Cephalosporines 5 000 500
Tetracyclins 5 000 50
A consequence for process development is that each product has its Insulin 8 125
Erythropoietin 0.01 500 000 000
own very specific production process due to differences in price and
volume, and because for each process different parameters are critical.
Interestingly, if we plot the (logarithmic) price of biotechnology products
against their (logarithmic) annual production volume we get more or
less a straight line (Figure 1.9). This typically tells us that the larger the
annual volume the lower is the price and vice versa. So, if we develop
a product for a large market the sales price will likely be low and the
techniques and raw materials that we select must be cheap.
On the other hand, if we have an exclusive product, it might initially
be interesting to be first on the market rather than to put a lot of effort
into the process optimisation to lower the price. This is characteristic of
protein pharmaceuticals, which are subjected to very high development
costs and development takes an average time to registration of 8-10 years.
The cost increases exponentially due to the expensive clinical trials that
are eventually performed by large studies on humans. In time, however,
competing pharmaceutical products will arrive and production volumes
may increase. In that case, price suddenly becomes important, and it can Figure 1.9: The price of common products
then be too late to think about process optimisation. This is especially produced by biotechnology in relation to the
annual volume. Very approximate data are
true if product quality resulting from the optimised process is different used. The arrow marks the general trend for
and requires retesting in clinical trials. This is a common dilemma for product development over time.
the pharmaceutical industry.
One example of the price development over time is the production of
penicillin. In the late 1940s the price of penicillin was very high due to
market shortage. Today, it is a bulk product with a price of approx 15
€/kg. There is thus a tendency with time for a product to move along
the straight line towards the lower right-hand corner in the figure as
indicated by the arrow (Figure 1.9).
The final cost of a product depends on many things including the
geographical localization of the production unit. If a product is cheap, the
costs of labour and raw materials are dominant. In such cases, companies
look for countries where such costs can be minimised. The effect of the
lowered prices on penicillium has for example meant that production in
Western Europe has been rapidly decreasing.
Figure 1.10: Development of penicillin unit-
s/ml per year 1940-1980. Developments were
made due to combined effects of: random muta-
genesis of strains, improvements in cultivation
technology and downstream processing and
through metabolic engineering
10 1 Introduction

1.5 Product categories

In order to obtain an overview of the market and to try to understand its


CELLS &
ENZYMES
characteristics, one can try to group products into categories. The basis
for this varies generally with the literature presenting it. Here, we divide
the bioproducts into four main categories (Figure 1.11) and we base this
division on the core technique used for product formation.
These four categories can be described further as:
1. Cells or cell constituents which are themselves the product or which
EXTRACED are used for various transformations
PRODUCTS 2. Extracted metabolites from a naturally derived source
3. Metabolites naturally accumulated as end-products in specific cells
4. Natural or unnatural (heterologuous) metabolites overproduced in the
cell by recombinant DNA technology
The first category includes the production of cells that are themselves, or
part of the cell is the product. These are often used to catalyse reactions
ACCUMULATION BY leading to products or various services. We shall look at some examples
ENVIRONMENTAL
MODIFICATIONS to understand this group into greater detail.
Living cells are often used to improve the quality of other products.
These are named “starter cultures” and include the production of the
yeast Saccharomyces cerevisiae to give taste and texture to bread in the
baking process. To this group lactic acid bacteria also belong, which
provide taste and acidification/preservation in sour milk production.
ACCUMULATION BY Cells are also used directly as a product such as “single cell protein”
GENETIC
MODIFICATIONS
used for food (e.g. Quorn) or cattle feed. Parts of cells such as plasmids,
i.e. non-chromosomal associated DNA used in gene therapy, which is
harvested from the cell, may also be categorised in this group.
Biotransformation using whole cells is used to perform reactions outside
Figure 1.11: Product categories. Products are
the cell and often involves various biotech services used by society. One
categorised according to the production tech-
nology and type of product example is the purification of wastewater, where excess organic carbon
and nitrogen are transformed into carbon dioxide and nitrogen gas which
can be recirculated back to the environment thereby closing carbon and
nitrogen balances in the nature. In addition to wastewater treatment,
biotransformation using whole cells includes also techniques for the
purification of soil and waste gases. Microorganisms can also be used in
consort with chemical reactions to release minerals in mining, particularly
in the case of low-grade mineral ores (microbial leaching).
Biotransformation using purified enzymes is a very large product cate-
gory with the production of high fructose syrup as a primary example.
Here, D-glucose is converted to D-fructose by the enzyme glucose iso-
merase bound to a solid support (HFS by Clinton Corn Products, 100 000
ton/year). HFS is used as a sweetener in many products.
The second category contains products extracted from a source in nature.
Many biopharmaceutical proteins were initially produced in this way
and some still are. One example is the extraction of human growth
hormone from blood which has now been replaced by recombinant DNA
production (see group 4 below) in the bacterium Escherichia coli or by
yeast. Strictly speaking, the extraction processes do not fulfil our
1.6 A comparison between bioproduction and organic synthesis 11

definition of a “Biotechnology”-based product since neither cells nor enzymes are used. However, if the
product in any way concerns the handling of proteins these processes are, by knowledge and tradition, often
developed and maintained by bioprocess engineers.
The third category includes the production of cellular metabolites, which are naturally accumulated as
end-products by particular cells and which can be enriched under specific environmental conditions both in
nature and in the laboratory/factory. These products originate from both primary and secondary metabolism
of the cell and the products range in size from small organic acids to large polymers. In the small molecule
area we find many examples of bulk chemicals such as ethanol, carboxylic acids and amino acids that are
primary metabolites but also biopharmaceuticals like penicillin which is an example of a secondary metabolite.
Microbial cells are also able to accumulate polymers or building blocks for polymers. Such polymers include:
nucleotides, polyesters, xanthane, hyaluronic acid and cellulose. Although we have found such cells in nature
that are capable of accumulating large amounts of these products, screening for mutants or cell engineering
is often used to increase the cellular yield and productivity of these products. This category constitutes the
largest and most diverse product category.
Often, the production in the original cell or extraction from a natural resource cannot provide the quantities
necessary for a competitive market price. This brings us to the fourth category of products where recombinant
DNA technology is used to introduce foreign genes and metabolic pathways into a microorganism. The
target cell for this modification can vary, but it might well be that the product is totally unknown for the
host cell. Examples of products produced by this technique are proteins such as: hormones, antibodies and
vaccines, all belonging to the pharmaceutical area. However, certain bio-chemicals and biomaterials, which
are not proteins, are also produced in this way. One example of a commercial full-scale process is the Du
Pont/Tate & Lyle production of the monomer 1,3-propanediol, which is produced by Escherichia coli by
expressing a novel heterologuous metabolic route leading to this product. After recovery, this monomer is
used for polymerization into polyesters for the production of biomaterials. The production of biomaterials, or
their building blocks, has become a growing research area in biotechnology, and there are many start-up
companies following in the wake of developments in currently ongoing research. The products of the 4th class
often aim to replace oil-based products or to create new products with interesting and desired properties.
Especially this category of products has developed rapidly over the past decade, with products including
farnesene, succininc acid, 1,4-butanediol and many others.

1.6 A comparison between bioproduction and organic synthesis

Products that are valuable to society can often be derived by the use of alternative production techniques. In
biotechnology these are based either on the cultivation of living cells or on enzyme transformation according
to our definition. The following positive benefits are often claimed:
I It is possible to use a range of waste streams from other industries as substrates
I Reactions can be performed at near room temperature and atmospheric pressure
I Cells or enzymes can perform many reactions in a single step
I Enzymes can perform stereoselective reactions
I It is possible to build very complex molecules often in a few steps
I Reactions can usually be performed with water as solvent
Bioproduction can often provide sustainable and environmentally friendly new products or alternatives to
existing products. However, bioproduction is not automatically or always the best choice. Processes might
result in byproducts, for instance carbon dioxide and heat evolution due to the metabolic activity. Depending
on the process technique it might also lead to the creation of nutrient- and energy-rich liquid effluents. All this
has to be taken into account, and companies have to either destroy it themselves by building and operating
purification units and/or by paying a fee to another company or to the community-operated wastewater
plants to take care of it. When for instance comparing bioproduction to organic synthesis, you can see that
bioproduction also has specific disadvantages, such as:
12 1 Introduction

I The use of very dilute systems requires access to large quantities of fresh water
I The reaction rates are comparatively slow to chemical synthesis
I The dilution requires large-volume equipment and handling
I Enzymes can perform stereoselective reactions
I Capture techniques of the product from very dilute systems needs to be added
I The investment costs are high for the biotech-based equipment due to the use of microorganisms which
require auxiliary equipment such as e.g. equipment for sterile operation

As always, no technology is beneficial in all respects, so a particular decision has to be taken in each specific
case without bias towards anything other than the final economy, safety and product quality.

1.7 The industrial bioprocess

The commercial production of all types of biotechnological products is generally divided into different but
very characteristic phases. These are, in turn, characterized by the use of different technologies, or by different
“unit operations”, which are also very well defined and which can be both of biochemical or chemical
engineering origin.
Generally we talk about three phases in bioprocess design: the upstream processing, the cultivation/biore-
action step (midstream) and the downstream processing where the latter involves the purification of the
product from interfering compounds and a simultaneous concentration of the liquid product. Apart from
these steps, the design of an additional number of subsequent steps must further follow before the product
can be released to the customer. These include the “formulation” and stabilisation of the product into the
desired customer-desired format but also the packing and distribution system. However, since these are not
primarily of biotechnology origin, they are not discussed in this book.
The design of each step of the process has a range of different techniques and components to choose from.
Since the development process is linear it should immediately be realised that constraints introduced in the
early stages will greatly reduce the degrees of freedom for the engineers operating further downstream. In a
company it is thus very important that all steps are considered more or less simultaneously.
In the upstream processing important choices are made with respect to the catalyst and the medium. The
catalyst is either a living cell or an enzyme. The biocatalyst is almost always optimized in one way or another.
For a living cell, this can include anything from screening for new microorganisms in nature to design of
new organisms with improved properties in the laboratory and also combinations thereof. The methodology
can be anything from large sets of random mutation experiments to the use of precise alterations based on
detailed metabolic knowledge and the use of recombinant DNA (rDNA) techniques. Upstream processing
can also involve the attachment of the cell or an enzyme to a solid support by immobilisation, which is used
to increase the catalyst stability and density and to be able to facilitate their separation after the product has
been synthesised.
The design of the media for the cultivation/bioreaction step need specific consideration and here the economy
is of considerable interest for many products. For some processes, these are difficult to design and some
companies provide commercial solutions for many types of cells and processes. Techniques for rational
optimisation of media by use of multivariate analysis are also commercially available. Raw materials are
carefully selected to fit the process with regard both to the requirements of the microorganism (catalysis
performed by the cell) or enzyme (catalysis performed by the enzyme) and to economic factors. Raw materials,
after being selected, need sometimes to be processed, especially if these are complex (e.g. polymers), since
most cells can only take up and use very simple monomeric nutrients. Media are also sterilized by a variety
of techniques before use. Medium design is further presented in chapter 3.
The reaction step, is a cultivation process if the catalyst is a living cell but if it is based on an enzyme reaction
we call this process “biocatalysis”. The reaction step is carefully designed and optimised with respect to the
type of bioreactor, the measurement and control of important variables in the bioprocess and the cultivation
technique used. The design is tested on a small laboratory scale before being scaled-up in consecutive steps
1.7 The industrial bioprocess 13

Raw material for medium

Propagation

Biocatalyst

Cultivation
Isolation & Purification

Figure 1.12: The main steps of an industrial bioprocess

to the final production scale. The design of this step is described in the remaining chapters throughout this
book.
The cultivation broth containing the microorganisms, the product and a surplus of medium components
is then harvested and the end product is isolated and purified to the desired quality. The product can be
localised either inside or outside the cell. The purification, or the “downstream processing - DSP”-steps,
may include a selection of different unit operations to purify the product to the degree needed such as: cell
homogenisation (if the product is inside the cell), sedimentation, filtration, crystallisation, precipitation,
extraction, adsorption and chromatography. The downstream processing is beyond the scope of this book
and will thus not be discussed further.
Is there a need to be specialised in all these techniques as an engineer? If we deal with a large company,
this is organised into different units, each with different goals, skills and purposes. Such a large producing
company has separate divisions for Research, Process development, Production and Quality control and the
employees are specialised in different subjects connected to the goals of each division. New products and
process design issues are solved in smaller or larger projects and these projects include participants with the
competence needed to solve a particular problem.
When a new product is developed it is vital for a company to be first on the market with the best product.
Parallel to the development of the product, the production process must also be designed. This process
includes all the elements described above.
The Research division is responsible for the development of the product and the first steps in product design
precedes all other steps. During this development small quantities of the product is needed which means
that the Research division will take the first steps also for the process development to get the appropriate
quantities for this first phase. This usually includes only the design of the production cell. When the product is
reasonably well designed, the cell and the initial production data are transferred to the Process development.
This division is generally further subdivided into groups for cultivation, downstream processing and product
analysis. During Process development, all unit operations of the process are settled, critical in- and outgoing
process parameters are investigated, the stability of the process to disturbances is evaluated and analysis
protocols are derived. Finally, the process is transferred to the Production unit for scale-up to the final
production scale, designed to meet the expected need of the company market share, and finally run and
controlled according to the preset requirements.
Production Cells 2
Carl Woese, who received the Crafoord Prize in Biosciences in 2003 2.1 Eukarya . . . . . . . . . . . . . 16
(the “Nobel prize” in Biology), was the person who originally suggested 2.2 Archaea . . . . . . . . . . . . . 18
the creation of a Universal Phylogenetic Tree for all known organisms. 2.3 Bacteria . . . . . . . . . . . . . 19
Gram negative bacteria . . . 19
The foundation was proposed to be the genealogical relationships of
Gram positive bacteria . . . . 20
the smaller subunit of ribosomal RNA. The tree was built from these
Cyanobacteria . . . . . . . . . 20
studies and the comparatively recent discovery of the characteristics of
the domain Archaea was primarily a result of this structural comparison.
We now consider organisms as divided into three domains: Bacteria,
Archaea and Eukarya.
Principally, this tree also describes all the possible production organisms
that we can use for biotechnological purposes. However, even though
we have discovered many new components produced by the multitude
of newly discovered cells, we know very little of their performance
under industrial conditions and some of them we cannot even cultivate.
Today we thus use only a very few of them for production purposes.
Furthermore, a commercial company tends to use only a very minor
subset of organisms of which they have a vast knowledge. Below follows
short descriptions of cell systems and their products that are used
commercially today.

EUKARYA
Animals
ARCHAEA
Slime
Entamobae molds Fungi
BACTERIA
Green nonsulfur Methanosarcina Plants
bacteria Methano- Extreme
Mitochondiron bacterium halophiles
Ciliates
Gram positive Methano-
bacteria coccus
Thermoproteus Thermo-
Proteobacteria plasma
Thermo- Flagellates
Chloroplast coccus
Pyrodictum Methano-
Cyanobacteria !4"#2#&10* pyrus
Marine Trichomonads
Flavobacteria Crenarchaeota
Microsporidia

Thermotaga Diplomonads

Thermodesulfobacterium

Aquifex

Figure 2.1: The Universal Phylogenetic Tree describing the relations between different cells. The tree is based on 16S or 18S RNA identification. Based on
a diagram in Woese CR, PNAS 97:8392-8396, 2000
16 2 Production Cells

2.1 Eukarya

Eukaryotic cells are either used as single cell preparations or as whole


organisms for production of products. One example of the latter is the
use of milking animals where the product – here often a pharmaceutical
protein - is genetically manipulated to appear in the milk. Although it
takes time for the production system to be designed and the requirement
that milking has to be continuously induced, the handling of animals
and the milk is a well-established procedure and the infrastructure is
thus already present which is a great benefit. We will not treat this type
of production further in the present context since it does not involve the
techniques of cell cultivation processes.
Animal cells are typically 20-100 𝜇m in size, which means that they are
comparatively large with respect to microorganisms. The cells grow by
cell division and although some can be grown in suspension, others are
dependent on anchorage to a solid support. Since they lack the cell wall
and since they are taken out of a multi-cellular environment, they are also
more difficult to grow as suspended cells than organisms that are evolved
for a unicellular existence. The different types of animal cells developed
today are used solely for pharmaceutical production and only when no
other cell type can fulfil the task. A characteristic example is the synthesis
of the complicated post-translational patterns of pharmaceutical proteins
such as glycosylation of human proteins.
The construction of hybrid cells of two mammalian cell types combining
two important properties – growth in suspension and the production of
antibodies – was realised by the construction of “hybridoma” cells. These
are lymphocytes fused to myeloma cells lines (cancer cells of the immune
system) and this achievement constituted a technology break-through for
the production of monoclonal antibodies and is today the state-of-the-art
for these products. There is currently also a great interest in the possibility
of growing undifferentiated stem cells for further use in medicine. This
is however still very difficult and the amounts that are possible to grow
are very low.
The drawbacks of animal cells are however many: the high cost and
difficult development of complex media, the expensive equipment, the
complex cell metabolism and the difficult and limited genetic manipula-
tion tools. It is also difficult to growth the cells in submersed cultures
other than for a limited number of generations and taken together with
the slow growth rate (the doubling time varies but it is often 12-25 hours)
this restricts the productivity limits. The latter is particularly trouble-
some since many products, such as antibodies, need to be produced in
large quantities. The handling of these cells requires therefore that the
personnel is skilled in a complex and broad scientific field and all these
factors lead to an expensive production system which can only be chosen
if the product has a very high value. Animal cells are thus used only
Figure 2.2: Antibodies can be produced in when no other system is applicable.
milking goats
Plants and plant cells are nowadays often explored as an alternative to
other production systems and also here, as with animal farming, the
technology of plant handling is already developed. The growth and
production of whole plants is however beyond the scope of this book
and will not be further discussed. Submersed plants cells are not used
2.1 Eukarya 17

to a larger extent for industrial production today. An exception is the


production of shikonic acid, which has been in operation for several
decades. The plant cell used is Lithosporum erythreum, a borage plant,
and the product has anti-tumour and wound-healing properties and
is also used as a pigment in lipsticks. Plant cells grow very slowly and
they are, like all the higher Eucarya, successively differentiated into
different cell types. The growth and development of leaves, roots, etc,
can however be steered in cultivation by the use of different chemicals
such as hormones.
Fungi and some unicellular green algae are sometimes called “eukaryotic
microorganisms” and they are used in many commercial production
processes and thus relate to many important product groups. The fungi
are further subdivided into: molds, mushrooms and yeast. Molds and
yeast are the common biotech producers. Most of these are aerobic
organisms.
The major product groups from molds are proteins (specifically enzymes),
organic acids and antibiotics. Common industrial molds are Aspergillus
niger (citric acid, 900 000 ton/year at a price of 1€/kg), Aspergillus oryzae
(bulk enzyme production at a price of 3€/kg), Penicillium chrysogenum
(penicillin, 6APA – a building block for semi-synthetic antibiotics), Peni-
cillium roqueforti (blue-veined cheese) and Fusarium venenatum (Quorn
protein). Commercial enzyme products from Aspergillus are: 𝛼 -amylase,
amyloglucosidase, proteases, pectinases, glucose isomerase, lipases, hemi-
cellulases and glucose oxidase. Mucor is also a common producer of
enzymes.
Mold production processes are started from propagation of an inoculum Figure 2.3: Top: Aspergillus niger grown on
of spores (conidia) that have been previously cultivated and harvested. an agar plate. As soon as the conidiophores
reach the air, they will form spores. Although
The spores, that are round in shape and the size of a small bacterium i.e. the mycelium is white, the aerial hyphae and
2 𝜇m, will after inoculum germinate and hyphae will start to protrude spores are often coloured which makes the
colony look dusty (moldy). Bottom: Hypha
from the spore. The hyphae will continue to grow and forms eventually
from Penicillium
a mycelium (mold) which is a dense network of hyphae.
The formation of a mycelium is a drawback in production processes since
it greatly increases the broth viscosity, and the mixing and aeration of the
bioreactor becomes costly and the process difficult to control. Molds can
produce very large amounts of products which are normally excreted into
the medium from which it can then be captured. The high concentration
of citric acid that can be accumulated in the environment by Aspergillus
tells us further that this cell is very acid-tolerant and has the ability to
keep its intracellular pH at a neutral level in spite of the accumulation of
acid.
The production capacity of enzymes from fungi is unparalleled amongst
other organisms and figures up to more than 100 g/L have been pub- Figure 2.4: Saccharomyces cerevisiae, bud-
lished. Fungi are by nature very good enzyme producers since they are ding
developed to excrete the enzyme for the purpose of degrading polymers
in their environment and in this way obtain new substrates for their own
growth.
Yeast also belongs to the Fungi, where the common commercial producer
is Saccharomyces cerevisiae. Yeast grows as single cells by a division process
that is called budding. A yeast cell is much smaller than the general
eukaryotic cell and the shape can be oval, spherical or cylindrical. The
18 2 Production Cells

yeast cell is however much larger than most microorganisms, approxi-


mately 5𝜇m in diameter. Its growth rate is slower than that of the average
microorganism but faster than that of molds and animal cells. Cells of
Saccharomyces cerevisiae are commercially used to produce baker’s yeast,
beer and ethanol but also for the production of biopharmaceuticals.
This yeast cell is a GRAS-organism (Generally Regarded As Safe) and
the genome has been sequenced. The products formed by S. cerevisiae
indicate that it is facultative aerobic and can thus grow both with and
without oxygen. Yeasts are also secretors like some other Fungi i.e. they
can export products e.g. proteins to the medium. The yeast Pichia pastoris
is a comparatively new organism in production processes and is only
used to produce a small selection of pharmaceutical proteins.
Algae, and particularly certain green algae, have gained a lot of in-
terest recently for their ability to fixate carbon dioxide from sunlight
(phototrophic) using water as electron-donor and in this way produce
interesting carbohydrates. Some species have also gained interest for
their ability to produce long-chain carbohydrates which approaches the
Figure 2.5: Green algae. Top: Volvox (unicel- consistency of crude oil which can be used for fuel production.
lular in colony). Bottom: Botrycoccus brauneii
colony which can produce fatty acids with a One example of such a green algae is Botrycoccus braunii which produces
chain length of up to 30-36 carbons carbohydrates with a carbon skeleton length of C30 -C36 . The size of the
algae cell varies considerably, from 5 to 100 𝜇m.

2.2 Archaea

The microorganisms belonging to this group have only recently been


recognised as promising for industrial production probably because
of their relatively recent discovery. Archaea have a prokaryotic cellular
structure but they share also several features with Eukarya. Many Archaea
live in the nature under very extreme conditions. Their environment
is often highly halophilic, acidophilic or characterized by very high
temperatures. Microorganisms that tolerate these conditions often colour
the environment vividly due to their production of pigments.
Extremely halophilic microorganisms (requiring a concentration of >
1.5M NaCl for growth) have been found to store polyesters, which uphold
the osmotic pressure of the cell. These compounds, which are variants
of the polyesters, polyhydroxyalkanoates (PHA), accumulate to high
densities in the cell (>80%) provided that a carbon source is available in
excess while some other nutrient is limiting. The three genes necessary for
production of the PHA´s have however also been cloned into recombinant
E. coli and also into plants.
Microorganisms living at high temperatures, the thermophiles, are also
very interesting in biotechnology. One reason is that the enzymes of
the cells that have been developed under extreme conditions might
exhibit unique properties that are interesting for various biotechnology
applications. Such enzymes have been used in many industrial processes
and have been the starting point for research on enzyme evolution in the
laboratory.
Microbial mining is a process where the reactions of specific acid-tolerant,
acidophilic bacteria and Archaea are used to complement the chemical
2.3 Bacteria 19

reactions that degrade mineral ore. The ability of these microorganisms


to use other electron-acceptors than oxygen for respiration (energy
generation) contributes to the success of this process.
Methanogens are Archaea that in a consorted action with other species,
called syntrophy, produce methane gas from the degradation of biomass
- not only in nature but also from industrial waste. They were first used
in connection with wastewater-treatment plants that are rich in biomass
as a waste product from this process. Today, both very large plants as
well as small household plants are being built to produce biogas as a
source of energy and a fuel.

2.3 Bacteria

The Bacteria contains many different phyla, and common microbiology


textbooks suggest that there are about 80 groups as based on 16S RNA
findings. Most of these still have not been cultivated and little is known
about their characteristics. Of commercial interest are today mainly the
phyla: Proteobacteria and Gram positive bacteria, which will be dealt with
into greater detail below.

Gram negative bacteria

The phylum “Proteobacteria” is the largest of the bacterial phyla and it has
also attracted considerable interest for commercial production and social
services. Proteobacteria are all Gram-negative, but they are otherwise
physiologically and morphologically very diverse. Several bacteria of
this group are used in consort e.g. in wastewater-treatment plants, where
the action of these microorganisms allows the closing of e.g. the carbon
and nitrogen cycles. To these belong e.g. the Nitrifying bacteria with
Nitrosomonas and Nitrobacter as common examples.
Acetic acid bacteria, such as Acetobacter and Gluconobacter are commonly
found in alcoholic fruit juices. Under specific conditions, these flagellated
cells carry out incomplete oxidation of alcohols and sugars leading to
an accumulation of organic acids as end products. Acetobacter are used
for aerobic acetic acid production and can grow well at low pH. Having
a complete citric acid cycle, these microorganisms can oxidize the acid
further to CO2 if sufficient oxygen is available which of course is unwanted
in the acetic acid production process. These cells can also oxidize sorbitol
to ascorbic acid which is an example of another biotransformation process.
Gluconobacter can be used to form e.g. gluconic acid from glucose due, in
this case, to an incomplete citric acid cycle.
Enteric bacteria, to which Escherichia, Salmonella and Enterobacter belong,
is a homogeneous group of facultative aerobic, non-sporulating rods.
Several of the non-pathogenic strains have attracted considerable com-
mercial interest because they offer the possibility of specific end-product
accumulation of valuable products.
Of particular interest is the enteric bacterium Escherichia coli. This small
rod-shaped bacterium with a length of 2 𝜇m has been frequently studied
since it is easily accessed and handled. It is used in research and in many Figure 2.6: Escherichia coli
20 2 Production Cells

commercial production processes. Products range from small molecules


like amino acids to complicated protein molecules for pharmaceutical use.
The prices of the products are very different, where the pharmaceutical
proteins can be produced in quantities up to 1 ton/year as in the case
of insulin at a cost of 125 €/kg. E. coli multiplies readily on all types of
commercial media by division into identical daughter cells (growth by
fission) at a rate which is unparalleled to other production organisms. E.
coli is a facultative aerobe and grows thus both with and without oxygen.
The growth rate is however much lower when no oxygen is present.

Gram positive bacteria

Figure 2.7: Corynebacterium. The dark dots Gram positive bacteria have for many thousands of years been used
are polyphosphate granules. Their generation in the production and conservation of food, particularly in sour milk
can be used for biological phosphate removal
in wastewater treatment
preparations. Different microorganisms such as Lactobacillus, Lactococcus
and Streptococcus, are used for the production of lactic acid. Neither of
these cells contains cytochromes and they can thus not perform oxidative
phosphorylation. Instead, substrate level phosphorylation is used and
this results in lactic acid formation. Some of the bacteria from this group
are able to form endospores that are difficult to remove by heat treatment.
Examples are Clostridia and Bacillus. The Clostridia genus is characterized
by anaerobic growth whereas Bacillus can grow both with and without
oxygen. Different types of Clostridia, e.g. Clostridium acetobutylicum can be
used to produce solvents such as acetone, butanol and propanol. Bacillus
Figure 2.8: Swiss Emmentaler cheese. Note
has also been the subject of many research projects during the last 20
the characteristic holes made from the CO2 years since Bacillus and Aspergillus are the two largest producers of bulk
production and containment during bacterial enzymes. These enzymes can in many cases be excreted to the medium,
growth
which is a large production benefit. From Bacillus we also get the oxygen-
sensitive equilibrium of the products acetoin and 2,3-butanediol.
Actinobacteria is an important and large phylum of production microor-
ganisms to which both Corynebacterium and Propionibacterium belong.
Corynebacterium glutamicum is a rod-shaped aerobic, non-motile microor-
ganism that forms “clubs” or V-shaped forms during normal growth.
The specifics of this particular cell division, by a so-called “snapping”
movement, gives the cell population a characteristic look as shown in
Figure 2.7. This microorganism has been extensively used to produce
amino acids, especially glutamate and lysine, where the production
volume of glutamate extended 1 000 000 tons/year at the end of the
last millennium. These two constitute the two largest annual production
volumes of amino acids in the world. However, tryptophan, threonine
and the branched-chain amino acids; isoleucine, leucine and valine can
be also produced by this bacterium.

Cyanobacteria

Cyanobacteria are a particular group of microorganisms that are claimed


to have had a major role in the early biological evolution of our planet since
Figure 2.9: Top: Synechococcus. Bottom: Toly-
pothrix sp. Cyanobacteria are sometimes called they might have been the origin of the oxygenation of the atmosphere
“blue-gree algae” which is a misnomer as this is due to their phototrophic nature. They are a large heterogenic group, in
not an algae but a bacterial species. Cyanobace- particular in relation to the morphology with various forms as we see
rial morphology is very broad and ranges from
unicellular (top) to filamentous (bottom). from Figure 2.9. They are one of the major phyla of Bacteria with distinct
2.3 Bacteria 21

relations to Gram positive bacteria although the cell wall is similar to the
Gram negative cells.
Cyanobacteria have been vividly discussed due to their water blooming
properties and their concomitant production of neurotoxins in particular
in the Baltic Sea. Today however, they are of major interest also as
producers of different types of biofuels.
Cultivation Media 3
The development of a biotechnological process in a commercial company 3.1 The metabolic starting point of
takes place more or less in parallel to the development of the product. medium design . . . . . . . . . . 24
When designing a process for a new product it is important to always 3.2 Medium categories . . . . . . 26
3.3 Medium design based on the el-
keep the final goal in mind. As an example, the substrate price and
ementary composition of a cell 27
quality needs to be appropriate and the chosen microorganism needs to
3.4 Substrates . . . . . . . . . . . . 27
be able to grow under the conditions desired for the rest of the process.
Substrates including macro ele-
During the early steps of product development, the goal is to provide ments . . . . . . . . . . . . . . . . 28
the researchers with modest amounts of product to satisfy the needs for Substrates including the micro-
further development and to promote a fast market introduction. Here, and trace-elements . . . . . . . . . 31
the first steps of the design of the production process are actually taken Water . . . . . . . . . . . . . . . 32
but from a very rudimentary perspective. Under such circumstances, the Control of culture pH . . . . 33
medium is selected only to give the best possible conditions for obtaining Calculation of a simple defined
small amounts of the product. Process economy is rarely the primary medium . . . . . . . . . . . . . . . 33
focus at this stage. Sterilisation of media and
medium components . . . . . . . 34
However, once the product development is sufficiently in progress, the
design of the large-scale process comes into focus. The goal is now shifted
towards the development of a process where both the product quantity
and quality will have to satisfy the market requirements and where the
economic aspects will be a dominating factor. For the time being we shall
in the present context forget about the quality since it is very specific for
each product, but we should remember that for the individual company
quality is always in focus and this applies also to the medium.
When the production cell has been chosen, it is time to select the specific
medium for the process (singular: medium, plural: media). A medium
needs to provide the best possible conditions for growth but also for
product formation. What a cell can do depends on inherited properties
or properties that we have introduced by genetic engineering, but what
it will do depends on the micro-environment i.e. on the medium. The
medium we eventually will select for the production depends on the
optimisation towards the economical output. Thus, after consulting
the genotype of the cell, which we always need to do into very great
detail, we design the medium depending on a large set of requirements
which reflect both the cell and the process performance. In summary, the
composition of the medium needs to:

I Support the desired rate of cell growth and the desired product
formation
I Reduce byproduct formation and deviations from standard growth
patterns
I Have a high turn-over of the medium components to products in
order to minimise the waste
I Match the cost of ingredients to the final product price
I Avoid the introduction of components that could compromise
the product quality by contamination and/or lead to complicated
purification issues
24 3 Cultivation Media

I Add, or avoid, components needed for the specific product to be formed

Many media design strategies initially based on the requirements of a cell in order to fast duplicate its mass
which is often a good starting point. In order to produce a product, on the other hand, we often have to
include additional considerations (last point above). This could involve the exclusion of one component from
the medium since it might be toxic to the cell or that the product might only form under strict limitation. The
latter type of components can often be one of the basic elements of the medium but may also be complex
compounds like vitamins. There are actually quite a large number of such examples in important bioprocesses
now running industrially, maybe even the majority. For other processes, it may be necessary to add a specific
compound that leads to the desired formation of product. One example is the addition of the side chain
of particular penicillins where the side chain is taken up by the cells and is incorporated into the product.
Another example is the addition of an inducer of recombinant protein production based on a chemically
induced promoter. The strategy of component exclusion/addition is however strictly product-dependent
which means that the appropriate conditions must be identified on a case-by-case basis.
Generally, we will select a medium that is as simple as possible and where the role of each component is very
well known. Which of the factors above that becomes dominant depends on the product. From an economic
point of view, the following rule applies: if the product is cheap, the medium must also be cheap! This means
that if we want to produce a low price product like baker’s yeast, we cannot select highly purified medium
substances. In such cases we often use raw materials that are considered to be a waste product by another
industry. In the Baker´s yeast process for example, we may use molasses which is a waste product from the
sugar-producing industry. On the other hand, if we are dealing with a high-price product like pharmaceutical
proteins produced by animal cells, the medium complexity and costs of specific components are allowed to
be high. Here, our goal is instead to avoid the introduction of medium impurities that could have an impact
on the product quality required for pharmaceutical production processes.
With the term “medium” we usually consider the solution of components initially prepared before inoculation
of the bioreactor but we might also include components fed to the process over time. This could be nutrients,
pH-controlling substances and also “antifoam” which is intermittently added to many processes to prevent
foaming. The latter is a surface-active compound that is not consumed by the cells and thus tends to
accumulate in the reactor. This is unwanted since it might give problems in the purification process but is
anyhow unavoidable in order to keep the cells inside of the reactor and must therefore be used.

3.1 The metabolic starting point of medium design

Since the medium design starts from the control of cell growth it may be a good idea to first briefly repeat
requirements for formation of new cells and thus the cell reason “d´être”.
It is important to notice that all intracellular metabolic events start with the uptake of the components we
provide from the medium and it is here that the control of the following cellular processes is managed.
Substrate uptake takes place through the action of specific transporters located in the cell membrane. These
consist of highly evolved proteins that are sometimes genetically and physically clustered into groups with a
common goal. There is at least one individual system for the uptake of most components, and many times
several, and the specific transporter that is preferably used differ with respect to the actual concentration of
the particular component. This means that not only the type of nutrient but also its concentration is of utmost
importance for the overall performance. In E. coli, a large number of such transporters are known and they are
present both in the outer and inner membrane although it is the latter set that is the most tightly regulated.
We need to know how this works in order to get the desired uptake rate since this rate is a prerequisite for
achieving a specific rate of growth meaning that the cells basically grow as fast as the substrate is taken up.
A large part of the understanding of the central cellular processes for generation of new cells can be given by
an overview of the carbon and energy metabolism. A reduced form of a carbon source, such as glucose, is
taken up from the medium by dedicated transporters, where-after this substrate passes through a series of
oxidation steps in what we call “the catabolism” and which concerns both the glycolysis ans the TCA cycle.
3.1 The metabolic starting point of medium design 25

The carbon-carrying metabolites generated in the catabolism (and of


course also a simultaneous range of other substances) are presented to the
cells as a range of 12 central building blocks which are used to build all
new cells, i.e. for “the anabolism”. The formation of these building blocks
allows however the cell only to partly fulfil the demand to multiply. The
carbon taken up can thus be found in the new cells but it is to almost equal
measures also used to generate the energy necessary to run this process
Figure 3.1: Principles of aerobic metabolism
which is the second requirement for growth. The energy generation is
which is one of the cornerstones in medium
linked to CO2 generation, which becomes a waste product of the carbon design
metabolism, i.e. carbon that the cell loose to the environment. We need
to take this into account when the carbon content of the medium is
designed.
These oxidative events of catabolism must of course be accompanied
by a simultaneous reduction to capture the generated electrons. In
microorganisms, the main electron acceptors are either nicotinamide
adenine dinucleotide (NAD+ ) or nicotinamide adenine dinucleotide
3´phosphate (NADP+ ) (and sometimes FAD+ ). It should be realised
that it is crucial that the produced NAD(P)+ has to be regenerated
otherwise the catabolism would come to a stop. This is achieved during
fully aerobic conditions from the practical energy generation where the
electrons from the reductant are finally accepted by oxygen and results
in the conservation of energy in the form of ATP. The simultaneous
production of water is the second waste product apart from carbon
dioxide.
In anaerobic processes, on the other hand, the energy conservation is to
a different extent replaced by other mechanisms. Alternative terminal
electron acceptors for anaerobically growing cells may be found in the
medium since this often contains inorganic substances. One example is
the use of nitrate that is used by e.g. species of Pseudomonas to close the
nitrogen cycle in wastewater treatment. One consequence of the inability
to use oxygen is that the amount of conserved energy is much smaller
in anaerobically growing organisms and they, consequently, often grow
slower. To compensate, many facultative aerobes may start to take up the
carbon source at a higher rate. This is a phenomenon that was observed
for the uptake of glucose by yeast by Louis Pasteur already in 1857 and is
consequently called “the Pasteur effect”. We now know that the difference
in the rate of the glycolytic flux, at the presence or absence of oxygen, is
due the regulation by ATP of the enzyme “phosphofructokinase” in the
glycolysis. For appropriate design of the medium, such effect must of
course be known.
Even if the environment is fully aerobic, the cell may choose to make
not only new biomass of the carbon taken up but also other things and
this may lead to excretion of cell-specific “products”. One example is the
“overflow metabolism” which is present in many types of cells and will be
discussed in a later chapter. This has largely to do with the concentration
of the medium components, as we shall see, and this is sometimes desired
and sometimes needs to be avoided. It is thus easy to understand that
the metabolic routes chosen by the cell depends on the medium in which
it grows and that this is thus first important to know and secondly a very
important design target.
26 3 Cultivation Media

3.2 Medium categories

A medium is a set of carefully selected raw material components that facilitate growth and/or product
formation. Each component is added to provide a “basic element” for growth, i.e. carbon, nitrogen, oxygen,
etc. The individual raw material component of the medium that is providing the basic elements is called a
“substrate”. An example of a substrate is glucose and the primary element we target by the use of glucose is
carbon.
Complete, ready-made media, can be bought under different brand names directly from the producer, and
this should guarantee that the cells can grow well and that the desired product is produced. The drawbacks
are that these media are often very expensive, the exact content is often not known or displayed (neither
components and quantities nor their origin) and the different batches can vary considerably. Sometimes these
media can contain components of animal origin that are undesirable. Animal cell cultivation media are often
of this “ready made” character.
To make the medium oneself and to get a good composition, the literature contains a large range of different
suggestions in the form of media “formulae”. These can often serve as a good starting point and the particular
names of such formulae are often referred to. They have thus become popular for various reasons and each
process engineer usually has his/her favourite basal medium for a specific cell type. This type of medium is
often used with appropriate additions. Examples of such formulae in the literature are LB, M9, 2*YT etc.
The LB-medium is the formula for “Lysogeny Broth” that was originally published by Guiseppe Bertani in
1951, and this broth is often used for the cultivation of e.g. E. coli. This is an example of a “rich medium” (see
below). Principally, different minerals are used as a base for this medium, as for many others, and to this
composition, organic compunds such as “tryptone”, vitamins and trace elements are added. In the original
formulation, Bertani also added glucose, but this has often been omitted in later descriptions. However, to
leave out glucose has important consequences for the metabolism, as we shall see. Most industrial producers
mix their media themselves, based on their own formulations and by purchase of the ingredients (raw
materials/substrates).
Commonly used names of media are:
I Defined medium
I Complex medium
I Minimal medium
I Rich medium
A Defined medium is what it says: all the components are known with regard to both identity and
concentration. This medium can contain a wide variety of molecules, but their presence and concentrations
are known.
A Complex medium is a medium that for various reasons contains a complex component. A complex
component is often of organic origin and one such example is “yeast extract” which is basically just a powder
of whole but dry yeast which thus contain a “complex” mixture of different substrates. It should thus contain
all the components necessary for growth since it has the composition of a natural cell. The components are
however rarely defined, either in type or in quantity, since the analysis is not easy to perform and is also
very costly. Different types of other “hydrolysates” of various organic sources, in particular proteins, are also
common complex components.
A Minimal medium is a medium which is based on very few components, usually only salts and a sugar.
It does not necessarily mean that these components are provided in small quantities - only that they are
the minimum range of components required for growth. A minimal medium is usually also a defined
medium since it is easy to measure the exact quantities of salts and many of the pure monosaccharides. Many
production cells can grow on this simple type of medium but some can not.
A Rich medium contains components that make it unnecessary for the cells to produce a range of intermediate
cellular substances themselves. In addition to a complex medium, also a defined medium can be of a rich
character. For instance, this can basically be a salt medium with known amounts of added amino acids.
3.3 Medium design based on the elementary composition of a cell 27

3.3 Medium design based on the elementary


composition of a cell
Table 3.1: Elemental composition of an E.
When we design a medium, we may start from our knowledge of the
coli cell. These are approximate values and for
general proportions of the elements building-up the living cell, since each cell type, grown under a specific set of
the creation of new cells is one of our goals. These proportions can conditions, the amounts will vary
be given from an elementary analysis of a cell. The character of the Element w% in the cell
major constituting elements are of course the same for cells of all living Carbon C 50
organisms on earth, but the exact elemental composition varies with Oxygen O 20
the type of cell and the environmental conditions. However, a list of Nitrogen N 14
generalised amounts of these compounds is often a sufficiently good, Hydrogen H 8
although an approximate starting point for the design. An example Phosphorous P 3
of such a list is shown in Table 3.1. For more careful design purposes, Sulfur S 1
and if it can be afforded, it might be a good idea to actually perform an Potassium K 1
elemental analysis of the specific production cell under selected operating Magnesium Mg 0.5
conditions. Such services are commercially available.
All the elements present in a cell are often subdivided into three major
groups which are the Macro-, Micro- and Trace elements.
Macro elements make up most of the cell mass and are: carbon (C),
nitrogen (N), phosphorous (P), hydrogen (H) and oxygen (O). The oxygen
that goes into the new biomass during growth comes generally from the
oxygen in the mixture of the substrate components, while the oxygen
needed for respiration comes from the air that is supplied to the bioreactor.
Carbon, phosphorous and oxygen are also needed for the catabolism and
the energy generation to take place, while nitrogen and sulfur enter first
in the anabolic processes i.e. for building of the macromolecules and
finally the new cells. The nitrogen carrier in the cell is glutamate and the
sulfur carrier is cysteine and both are taken up by specific cell membrane
based transporters.
Micro elements are the basal elements of a cell that are of intermediate
concentration. These are for example: sulfur (S), magnesium (Mg) and
potassium (K).
Trace elements are needed only in very low concentrations. Typical
trace elements are: cobalt (Co), iron (Fe), copper (Cu), calcium (Ca),
manganese (Mn), zink (Zn) and molybdenum (Mo). These are used
for various purposes by the cell but mainly as cofactors for various
enzymes. Common for all these elements is that they should be provided
in sufficiently large quantities to promote growth and product formation.
Too large amounts, on the other hand, will instead inhibit growth, so this
needs to be carefully sorted out. A consequence of the latter is that if very
high cell densities are needed (which is very often the case), medium
components that cannot be added from the beginning (most of them!)
need to be added during cultivation to promote further growth.

3.4 Substrates

When we know which elements the cell needs and in which proportions,
it is time to choose the substrates that can best provide these elements.
28 3 Cultivation Media

Substrates including macro elements

The macro-elements are of course very important since they make up most of the cell material. They can be
introduced in many ways and we shall consider them individually more closely. Salts of the desired elements
are often used and especially phosphate salts are often added to the solution for two purposes: to provide
phosphate and to increase the buffer capacity. An alternative buffering agent is calcium carbonate.
CARBON is added by a large variety of substrates and the choice depends on the product. Each carbon
source is more or less reduced i.e. will deliver more or less electrons and result in different degrees of energy
generated. The simplest and most well defined carbon substrates are monosaccharides and these are the
preferred choice by most cells. These substrates are often glucose or fructose. Some cells can also hydrolyse
disaccharides, which allow the use of sucrose, maltose (and sometimes also shorter maltodextrines) or
lactose. Since the monosaccharides are often too expensive for cheaper production processes, hydrolysates of
starch- or cellulose-based biomass or waste products of beet and sugar-cane-based sugar production are
often used. With regard to concentration, we need to keep in mind again that the cell uses carbon for two
purposes: to build new cells and to produce energy. Since the carbon is a dominant cost, we shall consider
the cheaper carbon substrates in more detail below. Some organisms, particularly fungi, can use longer chain
carbohydrates directly due to the presence of hydrolysing enzymes in the cell membrane or cell wall.
Molasses
Pure glucose and sucrose are rarely used in industrial production with the exception of high value products
such as biopharmaceuticals. Molasses, which is a waste product from sugar production, is the residual
component when most of the sucrose has been removed by crystallisation. It is a dark coloured viscous
solution containing 50-60% (w/v) of carbohydrates, 2% of nitrogen-containing substances and some vitamins
and minerals. The overall composition varies with country and batch and depends on whether sugar cane or
beet sugar is the raw material. Molasses are for instance used in the production of Baker’s yeast.
Malt extract
Extracts from malted barley can be concentrated to syrups and they follow the same preparation as that for
preparing malt for beer production. The composition varies and the carbohydrate content is approximately
90% by dry weight. The malt extract is a combination of sugars of various lengths but nitrogen compounds,
vitamins, proteins, peptides and amino acids are also present. Malt extract is often used in cultivations based
on filamentous fungi and yeasts.
Starch and dextrins
These polymers cannot be used directly by many organisms and they must first be hydrolysed. Only certain
amylase-producing organisms are able to use them directly, e.g. some fungi. Starch is a component of many
fruits and vegetables and waste potatoes constitute one such source.
sulfite waste liquor and lignocellulose
This substrate contains cellulose, hemicellulose and lignin and, depending on the pretreatment, also
degradation products of these. The substrate must undergo hydrolysis (using heat, pressure, chemicals,
enzymatic and combination of these) resulting in a mixture of six- and five-carbon sugars together with
metabolites of these such as furfural, hydroxymethyl furfural (HMF) and often also acetic acid and these are
all toxic to most production cells. In the hydrolysis process, it is often avoided to degrade the lignin since
it will add to the cell-toxic byproducts. Few cells can however utilize the 5-carbon sugars that stems from
the hemi-cellulose, which will be left in the medium, and this reduces the turnover of this substrate which
is not desirable. Some organisms, like some yeasts and molds, are able to metabolise a part of these toxic
compounds even if not to completion. Some specialized microorganisms, like Clostridium thermocellum can
directly break down the cellulose part of the lignocellulose, whitout the need for pretreatment.
Whey
Whey is a waste product from cheese making. The annual world production is over 800 million tons. It
contains mostly lactose and milk protein. The material is difficult and thus expensive to store and transport
since it quickly spoils. Not all organisms can ferment lactose, and this makes it less interesting to use.
3.4 Substrates 29

NITROGEN can be added to the medium by many types of substrates.


Some nitrogen sources like ammonia (NH3 ) and the organic compound
urea or carbamide, (CO[NH2 ]2 ), can enter the cell simply by diffusion,
but urea is rarely found in any laboratory medium, although it may
in the future be an accessible and cheap sustainable nitrogen source.
The simplest type of substrate for nitrogen supply is a nitrogen salt
(ammonium hydroxide, ammonium sulfate, ammonium chloride or
ammonium nitrate). In this case, ammonium, NH4 + , is assimilated by
the cell through specific uptake systems, and enters the metabolism by
reductive amination of 𝛼 -ketoglutarate which results in the formation
of glutamate, a common intermediate in nitrogen metabolism in most
microorganisms. At the same time, the hydrogen ion has to be excreted
and this will continuously lower the pH of the medium which is important
to handle.
A common cheap complex nitrogen source is “Corn steep liquor” (CSL),
a waste product from the production of starch from corn wet milling. CSL
is a very viscous solution that during storage separates into two layers,
one clear brown top layer and one much thicker and light brown more
dense bottom layer. These layers have to be well mixed before the desired
amount is taken, in order to ensure a reproducible take-out and thus give
a reproducible process. CSL contains also vitamins and minerals and has
a solids content of about 50% (w/w).
Peptone and tryptone are derived from milk or meat protein by prote-
olytic digestion. In addition to small peptides, the resulting spray-dried
material includes fats, metals, salts, vitamins and many other biologi-
cal compounds. Tryptone is an assortment of peptides formed by the
digestion of casein.
Table 3.2: Approximate composition of a yeast
Yeast extract or yeast hydrolysates are essentially a powder of dried yeast extract
and their composition is therefore complex. Due to the complexity, it
Component % w/v
will provide most of the nutrients that are desired by the cell, but often
Protein, peptides, amino 73-75
the intention is primarily to add a source of nitrogen and vitamins. A
acids
standard composition example is shown in Table 3.2.
Free amino acids 35-40
Nitrogen can also be provided by the addition of amino acids which Peptides < 600 Da 10-15
are readily assimilated by the cell and which are preferred to other Material > 600 Da 20-30
nitrogen sources since it saves energy for the cell when they do not have Vitamins 𝜇g/g
to synthesise them themselves. This addition is either made as a pure Thiamine 30
amino acid cocktail or by a complex and much cheaper hydrolysate from Riboflavin 120
a protein-rich source. Examples of the latter are hydrolysates of fish, soy Niacin 700
or peanut meal, whey, casein or meat (see above). Bacteria like E. coli can Pyridoxine 20
make their own amino acids and it is basically not necessary to provide Folic acid 30
any of them. This is not the case for cells from higher organisms. Calcium pantothenate 300
Biotin 2.5
The use of a pure and defined amino acid mixture is expensive and
the design is never easy or straightforward. The reason is the very
tight control of amino acid metabolism. Amino acid production in a
cell takes place in a group-wise fashion and a set of amino acids share
thus a common start-up metabolic route which later on is branched-off
to the individual acids. Since the initial reaction sequence is feedback
controlled the addition of one acid might lead to the inhibition of the
production of all acids in the particular group. One example is the
branched metabolism of valine and isoleucine where the addition of
valine will inhibit isoleucine synthesis in E. coli.
30 3 Cultivation Media

ALA GLU THR ASP ALA GLU THR ASP


5 14 5 14
TRP PRO SER NH3 TRP PRO SER NH3

12 12

Ammonium Concentration (mM)


Amino Acid Concentration (mM)

4 4

10 10

3 3
8 8

6 6
2 2

4 4

1 1
2 2

0 0 0 0
0 2 3 4 0 2 3 4 5
Cultivation Time (h) Cultivation Time (h)

Figure 3.2: Consumption of the amino acids used for energy generation and the corresponding excretion of ammonium ions with cultivation time.
Left: LB medium with glucose. Right: LB medium without glucose. Amino acid bar chart from left to right: alanine, glutamine, threonine, asparagine,
tryptophane, proline and serine i.e. the so called “energy amino acids” i.e. those than can replace other more common carbon sources like glucose for
energy formation

Amino acids can be used by the cell as a source, not only of nitrogen and incorporated into the proteins of the
cell (growth), but also of carbon and energy. This is the reason why the growth rate will be considerably
higher when these acids are added.
Some of the amino acids can be used for energy generation if there are no other, and better, sources available,
such as glucose. This effect is shown in Figure 3.2 for an amino acid rich LB-medium with glucose (left graph)
and without glucose (right graph). When glucose is not present, a larger amount of these particular amino
acids is taken up and this is seen by the greater accumulation of ammonium (right graph).
The uptake of amino acids as a carbon source leads thus to a general increase in the pH of the medium. This
is both due to the removal of the amino acid from the medium but also due to the cell metabolism where the
intracellular oxidative deamination leads to removal of the ammonium group which, in microorganisms and
also in sea living animals like fish, is directly excreted to the environment. From the ammonium accumulation
seen in the figure it is obvious that there will be a higher increase in pH if glucose is not present. We have
LEU LYS+PHE ARG LEU LYS+PHE ARG
7 7
ILE CYS+VAL GLY ILE CYS+VAL GLY
MET HIS TYR MET HIS TYR
6 6
Aminoacid Concentration (mM)

5 5

4 4

3 3

2 2

1 1

0 0
0 2 3 4 0 2 3 4 5
Cultivation Time(h) Cultivation Time(h)

Figure 3.3: Consumption of amino acids used for incorporation only into the cell mass, as a function of cultivation time. Left: LB-medium with glucose.
Right: LB-medium without glucose. Amino acid bar chart from left to right: leucine, lysine+phenylalanine, arginine, isoleucine, cysteine+valine,
glycine, methionine, histidine and tyrosine

already stated that it is very difficult to design the medium with respect to the amount of the individual
amino acids and with an unspecific protein hydrolysate it is of course even more difficult. What adds to the
complexity of this design is that the depletion of one single amino acid can lead to a metabolic response which
may stop growth until the cell can adjust to this fact – if it can do so at all. E. coli cells possess a general
3.4 Substrates 31

basic control mechanism called the “stringent response” which will


affect, at least temporarily, some key features of the cell metabolism such
as protein transcription and translation. This is of course particularly
troublesome if amino acids are used also as a source for carbon/energy,
i.e. if there is no other source for this in the medium.
Some amino acids are however used by the cells only for biomass (protein)
and not for energy production. A comparison of the consumption pattern
of these particular amino acids is shown in Figure 3.3. The consumption
pattern is in this case obviously very similar between growth with and
without addition of glucose but the use of amino acids is slightly higher
when no glucose is present (right graph).
PHOSPHOROUS is assimilated by most cells as inorganic phosphate
and simple salts are therefore used in the medium, especially variants
of potassium salts. Phosphate is present in the cell in many forms both
as inorganic (orthophosphate Pi , pyrophosphate PPi and polyPi ) and in
numerous organic phosphates. Many of these have important regulatory
roles in the cell and a certain amount is during many conditions present
as a “free” pool. The major cell components containing phosphate are;
DNA, RNA and polyPi . Phosphate serves two purposes in the medium;
it is needed as a component in the cell building blocks and in energy
generation but it can also be needed in the medium to sustain the buffer
capacity. Under conditions of phosphate limitation in the medium, many
cells can form periplasmic or extracellular phosphatases to hydrolyse
organic phosphate present in the medium. Phosphate is thereby liberated
and can be taken up by the cell.
HYDROGEN and OXYGEN. Aside from the medium components, also
the water used to dilute and dissolve the medium and the substrate
used to provide the macro-elements simultaneously provide an excess of
hydrogen that is needed for incorporation into new cells. We need not
therefore consider adding this separately.
The air that is introduced into the bioreactor provides the necessary
oxygen for respiration. The oxygen incorporated into the biomass will
be provided in excess by the substrate components added to the liquid
medium in the same way as hydrogen.

Substrates including the micro- and trace-elements


Table 3.3: Trace element stock solution used
Micro and trace elements are often added as simple inorganic salts. in an amount of 1 mL/L medium. The EDTA
solution is added to avoid precipitation
Inorganic sulfur is largely present as sulfate and sulfide in the lithosphere,
and less abundant forms are thiosulfates (S2 O3 2- ), thionates (Sn O6 2- ) and Element Substrate Conc. (g/L)
elemental sulfur. sulfate and its derivatives, such as thiosulfate, are Fe FeCl3 *6H2 O 16.7
however the main forms of inorganic sulfur assimilated by bacteria and, Ca CaCl2 *2H2 O 0.5
for this reason, sulfate is the most common source of sulfur in laboratory Mn MnSO4 *H2 O 0.11
media. The cell rapidly reduces the sulfur to sulfite which is incorporated Zn ZnSO4 *7H2 O 0.18
into the metabolites. Magnesium is often added as magnesium sulfate Cu CuSO4 *5H2 O 0.16
and thus satisfies the need for both magnesium and sulfur. Co CoCl2 *6H2 O 0.18
Na2 -EDTA 20.1
In addition to the trace elements listed in Table 3.3, a balanced concentra-
tion of sodium and potassium is always needed to sustain the membrane
potential and transport processes there over. The need to provide sodium
for bacteria has not been altogether clarified but it is well established
32 3 Cultivation Media

for eukaryotic cells. Trace elements are generally prepared in a stock solution like the one showed in the
table. This particular example is used for E. coli and 1 mL/L medium will support the growth of up to
approximately 10 g/L of cells. If complex substrates are used in the medium this often means that a number of
trace elements are also simultaneously provided. Trace elements can also be added with the water depending
on the source.
Some cells need vitamins for growth and it is therefore necessary to add these to the medium. They are
added either in a pure form or as parts of a complex substance such as yeast extract. Processes where vitamins
are needed are generally based on yeasts (B-vitamins) and some fungi. Most bacteria and fungal cells can
however grow on very simple media containing merely salts and a sugar source for the supply of carbon. Cells
of higher organisms, on the other hand, that have been taken out of a highly controlled micro-environment,
such as a cell from a plant or an animal, need the addition of many specific components. These media are
thus very rich and also contain a larger number of components than the media for microorganisms including
a range of organic substances such as vitamins, unsaturated fatty acids, amino acids, hormones and other
growth factors.

Water

Almost all cell cultivations require large quantities of fresh water and this is very important to consider
especially for cultivation of cells in countries that do not have access to large fresh water resources. The purity
required depends however on the process. For cheap products in Sweden, the municipal water quality is quite
sufficient. This has also the advantage that some of the salts necessary for the production are added with the
water. For products used for human health such as pharmaceuticals on the other hand, highly purified water
(WFI-Water For Injection) is used to avoid the addition of impurities and thus to maintain product quality.
It is important that water is added to the medium in sufficient quantity to ensure growth. The concept of
“water activity” is here essential. For media with low molecular weight components, the water activity can be
approximately expressed in terms of the moles of water, nw , and the moles of solute, ns , as:

Definition 3.4.1 Water activity


𝑛𝑤
𝑎𝑤 = (3.1)
𝑛𝑤 + 𝑛 𝑠

The activity of pure water is therefore equal to 1. However, also human blood and seawater are very close to
unity being 0.995 and 0.980, respectively, and this is also the condition where most Gram-negative thrive.
Gram-postive rods and especially cocci can grow at lower values such as 0.90-0.95 and some yeasts at a value
down to 0.80-0.85. Halobacteria in salt lakes can grow at water activities down to 0.75.
A reduction in water content will mean that the cells cannot grow or divide and in the worst case the normal
flow of water into the cell (due to the accumulation of osmotically active substances in the cell) will cease and
the cell cannot keep up the hydrostatic pressure i.e. the turgor. Why is the water activity important? The
reason is the multitude of functions that water has: it is the solvent for all the molecules of the cell allowing
them to pass in and out over the membrane (transport mechanisms), it promotes the stability of molecules
through hydrogen bonding and also helps in the positioning of atoms (molecular interaction), it forces
non-polar substances to aggregate (formation of organelles such as membranes) and its polar nature makes it
cohesive by creating surface tension and a high specific heat. The water content is actually so important that
a reduction, by for instance drying or the addition of salt, can be used to preserve of food against microbial
activity (both the mechanisms of infection and of microbial degradation).
Since the medium components are consumed during growth, the microenvironment of the cells in many
cultivation processes is continuously changed. This means that the water activity and also the osmolality may
change and cannot be considered as constant. For such events, cells have developed efficient mechanisms to
protect themselves. An upshift in the osmolality of the medium will for example lead to the accumulation
of so called “compatible solutes”, and examples of such metabolites are; betaine, glutamine, ectoine and
3.4 Substrates 33

trehalose. If the osmolarity cannot be controlled by production of these substances, this can induce stress in
the cell which may or may not affect the product formation. In the production of Baker´s yeast, trehalose
accumulation in the cell is induced by deliberate changes in the medium because this substance protects the
cell from dying during storage under dehydrated conditions. Hyperosmotic conditions frequently occur in
cultivation due to a high concentration of substrates and products. This may lead to an increased need for
energy to sustain viability and this may, in turn, affect both product formation and product quality.

Control of culture pH

The pH of the cultivation is an important factor controlling the growth rate but many times also the product
formation and the product quality. This means that the pH must preferably both be measured and controlled
and which is always the case in industrial production processes. Standard and sterilisable electrodes are
used in the bioreactor and they are inserted and calibrated before sterilisation. Usually, the pH is checked
after sterilisation, since it is often slightly changed during this procedure. A sample is taken of the culture
medium and checked by comparative analysis on an external pH-meter.
The change in pH will depend on the cell metabolism and on the medium used as we have seen in this
chapter. It may either increase or decrease in the cultivation and some times an early part of the process goes
in one direction while a later part goes in another. In the latter case, both acidic and alkaline solutions need
to be prepared and used to keep the pH constant. This is usually standard in the fully automated bioreactors
provided today. However, we still need to choose the actual solutions for this purpose.
For E.coli grown in ammonium salts media at high glucose concentration, the pH will go down. Titration can
then be performed with sodium hydroxide and 1-3 M solutions are generally used. The concentration needs
however to be optimised in relation to the setting of the PID-parameters of the control equipment in order to
arrive at a low variation in the pH. The buffer capacity of the medium will here also play a role since it will
affect the rate of change of the pH. For high cell density cultivations, a common strategy is however to use
ammonium hydroxide for titration since this will also lead to a continuous nitrogen supply. The acids used
for pH control are often sulfuric or phosphoric acid. In animal cell cultivation, pH is often controlled by CO2
sparging.

Calculation of a simple defined medium

We have now summarised common medium components and how they can be chosen. We now come to the
calculation of how much substrate should be added? We can arrive at an approximation by again starting
from the elemental composition of a cell. That is: how much C, N, O, P etc does our production cell contain?
Many such tables can be found in the literature and the cell composition varies with organism and cultivation
conditions, as we noted before.
Let us therefore return to Table 3.1 with the E.coli elemental composition. If we make some assumptions
regarding the growth, we can use these proportions to calculate the quantities needed. The assumptions we
need to make, at least initially, are the following:

I We are only making new cells on the basis of these substrates. This is of course not true for carbon,
which is required both for new cells and for energy metabolism. For aerobic respiratory growth we can
estimate that half the carbon supply will go to new cells. This means that we have to take twice the
amount of substrate for the supply of carbon i.e. the theoretical need for this element is twice as much
due to carbon dioxide formation.
I All components can be taken up simultaneously
I We do not take into account the buffer capacity needed

Since we know from the literature how much of the elements are needed, and if we decide to use a set
of particular substrates, the amounts of the substrate components required to obtain 1 g of cells can be
calculated.
34 3 Cultivation Media

Table 3.4: Amounts needed to obtain 1 g of cells. *Since we use phosphate also to give the medium a certain buffer capacity, we need much more
phosphate in reality

Element needed Element in


Element Substrate Substrate needed (g)
(% of 1 g) substrate (w%)
C 50∗2 Glucose 40 2.5
O - Surplus in medium - -
N 14 NH4 Cl 26.17 0.54
H - Surplus in water - -
P 3 KH2 PO4 22.76 0.13*
S 1 MgSO4 ·7H2 O 13.01 0.077
K 1 KH2 PO4 28.73 0.035
Mg 0.5 MgSO4 ·7H2 O 9.87 0.05

Example 3.4.1 1 g of cells consist of roughly 14% Nitrogen (see Table 3.1). The substrate used as nitrogen
source is ammonium chloride (NH4 Cl), of which Nitrogen is 26.17 w%. The amount of ammonium chloride
needed for 1 gram of cells is:

𝑔𝑁
h i
𝑓 𝑟 𝑎𝑐𝑡𝑖𝑜𝑛 𝑜 𝑓 𝑒 𝑙𝑒𝑚𝑒𝑛𝑡 𝑖𝑛 𝑏𝑖𝑜𝑚𝑎𝑠𝑠 1 𝑔𝑥 ∗ 14
100 𝑔𝑥
= = 0.54 𝑔
𝑓 𝑟 𝑎𝑐𝑡𝑖𝑜𝑛 𝑜 𝑓 𝑒 𝑙𝑒𝑚𝑒𝑛𝑡 𝑖𝑛 𝑠𝑢𝑏𝑠𝑡𝑟 𝑎𝑡𝑒 𝑔𝑁
h i
26.17
100 𝑔𝑠

Sterilisation of media and medium components

Before the medium can be inoculated with the production strain, it needs to be sterilised to avoid contamination
by other organisms. Otherwise, if these grow more rapidly than the production strain, they may entirely take
over the cultivation. An infection will in any case lead to a risk of contamination of the final product and
impair the quality. It should however be stated here that not all processes can afford nor need the sterilisation
procedure because of the additional cost and other measures to keep the purity are employed.
The most common method of sterilisation is the application of heat. The medium is then often sterilised at
121°C at 1 atm overpressure. The time required for sterilisation depends on many factors, but it can usually
be calculated from the inactivation kinetics of contaminating organisms that are generally given in the
literature. In production, particular attention must be given to the prevention of contamination by sporulating
microorganisms where the spores are present in the air everywhere and by bacteriophages that can also be
very serious contaminants. Endospores are the most difficult ones to remove due to their thick cell wall, so if
the criteria for removal of these are fulfilled (time and temperature) all other contaminating organisms are
also removed and it is on this basis that sterilisation times are generally been calculated. It should be noted
at this point that the chemicals present in the medium may certainly also be affected by the sterilisation
procedure and unwanted reactions may appear.
The practical heat sterilisation procedure differs depending on the scale of the cultivation. In small batch
processes, the sterilisation is also a batch-wise procedure, whereas on a large industrial scale sterilisation
is often a continuous process. The medium can be sterilised within the reactor i.e. at the same time as the
reactor is sterilised or outside the reactor where after it is aseptically transferred to the sterile (or sometimes
only cleaned) reactor. The medium sterilised in a continuous setup is subjected to a very high temperature for
a very short time, often only a few seconds. This gives often a positive influence on the relationship between
the increase in the rate of chemical transformation of the medium components and the inactivation of
3.4 Substrates 35

microorganisms since the rate constant for spore inactivation is affected


much more than the chemical reaction rate constant by the increase in
temperature.
The sterilisation protocol for the medium is one good example of the
many types of process documents which are used by the industry to
supervise individual processes so that they will reproducibly give the
same product output every time they are run (amount and quality).
Taken all the documentation together, these ensure a traceable line for
each production batch, i.e. all the way from the strain origin and propa-
gation, the specifications of the medium ingredients, the addition and
sterilisation protocols, through all the synthesis and purification steps
to the packing of the final product. For many products the regulatory
authorities require this detailed procedure and its documentation to
allow the company production of the product.
Many medium components are negatively affected by heat which may be
emphasised when sterilised together with other components. Vitamins
and trace elements are thus often sterilised by filtration and added to the
medium after sterilisation of the main components. Some components
are sensitive to heat because heat increases the rate of chemical side
reactions. One such example is the Maillard reaction, which involves
the reaction between reducing sugars and substrates containing amino
groups. Such Maillard products may be toxic to the cell and may cause a
discoloration of the product (yellow to strong brown colour). Glucose,
a common carbon source in many processes, is thus always sterilised
separately (by heat) and added aseptically to the medium after all the
other components have been sterilised.
Membrane filters can be, and often are, used for sterilisation, as discussed
above. The membrane filters are also called absolute filters since the pore
size of the filter is defined and gives a cut-off at a specific size of the
penetrating molecule
For cell removal this size is between 0.2 and 0.45 𝜇m and it excludes
microorganisms from passage. Membranes are typically made of ny-
lon, polyvinylidene fluoride (PVDF) or polytetrafluoroethylene (PTFE).
Membrane filters should be distinguished from “deep filters” which
Figure 3.4: Membrane or absolute filters for
remove microorganisms depending on the depth of the filter and are sterile filtration
consequently only able to remove a specific amount of contaminant.
Such filter materials are typically made of glass fibre and polypropylene.
When cultivation is performed in shake flasks, cotton stoppers are often
used as air filters. These are deep filters and their depth and character
thus determines the reduction of the number of microorganisms that
can penetrate. A precaution using these cotton stoppers needs to be
taken since wetting of the cotton will effectively stop oxygen transfer
to the liquid. It should be noted that sterilisation in bioprocesses is not
only important for the medium, but also for buffers and cultivation and
post-cultivation additives as well as for the ingoing air.
Growth of Cells 4
We have now looked at a range of different microorganisms and have 4.1 Growth forms . . . . . . . . . 37
discussed how to supply these with the components essential for growth. 4.2 Storage of production cells . 39
If we then take our designed medium and add the microorganisms, i.e. if 4.3 Inoculum preparation . . . . . 41
4.4 Growth patterns . . . . . . . . 42
we inoculate the medium, what will happen? Well, all types of organisms
The phases of a growth curve 42
will start to grow, provided all required components are present, and the
4.5 The rate of growth . . . . . . 43
cells grow according to a pattern characteristic of the particular cell type
4.6 Influence on the growth curve of
and the given environment. dead and/or non-viable cells . . 46
4.7 Factors affecting the maximum
growth rate . . . . . . . . . . . . . 47
4.1 Growth forms 4.8 The time-dependent uptake of
sugars . . . . . . . . . . . . . . . . 49
4.9 The substrate uptake rate as
Bacteria and animal cells grow by division into two identical daughter a function of substrate concentra-
cells, a process called “binary fission”. This process is well known and tion . . . . . . . . . . . . . . . . . . . 51
is described into some detail in standard microbiology textbooks. This 4.10 The Monod equation . . . . . 52
propagation pattern includes many of the production organisms of 4.11 Batch growth with limiting
interest to us. The equal distribution of the material from one cell into substrates other than the carbon
two new cells means that we cannot e.g. say that one cell is older than source . . . . . . . . . . . . . . . . 55
the other - there is no old mother and no new daughter. 4.12 Growth on mixed carbon or
mixed nitrogen substrates . . . 57
In principle, the bacterial cell will follow the growth cycle events and 4.13 Cellular production of in-
continue to grow as long as we provide it with fresh medium. When hibitory substances - overflow
animal cells grow and multiply by division, it is however found that only metabolism . . . . . . . . . . . . . 58
a very limited number of generations can be generated after inoculation 4.14 Techniques to measure growth of
in spite of the medium. The reason for this is not clear but this type cells . . . . . . . . . . . . . . . . . . 60
of cell is of course not evolved for growth in liquid suspension, as are Direct counting of cells . . . . 61
bacteria. Despite our efforts to design a suitable medium the cells will Light scattering . . . . . . . . . 61
Weighing . . . . . . . . . . . . 62
most likely suffer from lack of the control measures taken by the host
4.15 Measurements of medium and
organism from which they are extracted. Research has also indicated
intracellular components . . . . 63
that these cells make e.g. their own growth inhibitors and also growth
promoting substances and this further complicates the issue. Whether
this is a continuous process or whether it starts after a certain number
of generations is still not known. Re-inoculation into a fresh medium,
without such factors present, seems to lead to a continuation of growth
in some cases. Growth of animal cells follows the general growth cycle
events but occasionally, and particularly when used for production, a
temporary arrest in selected phases of this cycle is seen. The reason why
these cells eventually die may thus have nothing to do with the medium.
The animal cell dies by either of two mechanisms: by necrosis or by
apoptosis. These mechanisms are very different. Necrosis is a rapidly
induced state leading to cell death and leakage of the cytoplasmic content
to the medium and may be a result of e.g. stressful conditions during
bioreactor cultivation. Apoptosis, or programmed cell death, is a longer
process where the cell is reorganised into apoptotic bodies where the cell
content is stuffed into smaller membrane-bound vesicles. In the human
body they are removed by phagocytosis.
38 4 Growth of Cells

Figure 4.1: Exponential growth of a mycelium from one initial spore

Fungi on the other hand, are often subjected to a very different growth process that results in the development
of specific cell morphology. Molds thus grow by a mechanism called “apical growth” - a growth process that
starts with inoculation of a spore, which upon suspension in a liquid medium starts to swell and increases in
size. Eventually a small pointed shape, a “bud”, protrudes from a part of the spore (a germ tube), i.e. the
spore germinates. This bud is extended into a hypha (plural: hyphae) which grows in length but only from
the tip, hence the expression “apical growth” (Figure 4.2). The rate is initially exponential but as growth
proceeds it becomes linear. The length of a hypha is indeterminable but it has a fairly constant diameter that
can vary between 2 and 30 𝜇m.
The hyphae will at a certain point branch and the branches will grow, also from the tip. Growth continues
in this way until a large “tree form” is created and we have a mycelium (plural: mycelia)(Figure 4.1). The
continuous formation of new tips and the fact that this leads to new branches supports the exponential
growth pattern even though the hyphae themselves grow in a linear fashion. The mycelium can further form
dense clumps of mycelia called “pellets”. The degree of pellet formation compared to dispersed mycelium is
dependent of a large variety of inoculum and process conditions. In conclusion, the morphology of this type
of cell is thus very different from that of common bacteria.

Figure 4.2: Principal drawing of the morphology of a fungal hypha

When the tips grow and a critical amount of protoplasm is formed, the central nucleus divides and a septum
(plural: septa) is created at the point of division, creating “individual” cells, each with one nucleus (Figure
4.2). The septum is a cross-wall and it supports the shape of the hypha. The tip continues to grow and divide,
but this division takes place only after a certain length of the hypha has grown out or, to put it in another way,
when a certain cytoplasmic volume is formed in connection to an actively growing tip. This means that there
should be a clear relationship between cytoplasmic volume, nuclear division and branching. It has indeed
been shown that if we allow spore germination and hyphal growth until exponential phase is reached, a
certain constant “length of a hyphal growth unit (G)”
4.2 Storage of production cells 39

could be determined:

𝑡𝑜𝑡 𝑎𝑙 𝑙𝑒𝑛 𝑔𝑡 ℎ 𝑜 𝑓 𝑚 𝑦𝑐𝑒 𝑙𝑖𝑢𝑚


𝐺= (4.1)
𝑛𝑢𝑚𝑏𝑒𝑟 𝑜 𝑓 ℎ 𝑦𝑝𝑎𝑙 𝑡𝑖𝑝𝑠

This length varies between 30 and 400 𝜇m. When the constant volume is
established, the mold is growing at its maximum growth rate which can
then be determined. It is however very impractical to count cells of molds
and instead the weight of the mycelium is used for this purpose. The
increase in weight follows a pattern that is identical to e.g. that of bacterial
growth. In contrast to bacterial growth, however, the mycelial cells are
not identical but have different ages. The newest cells are always at the
tip and contain certain specific cell constituents, whereas further back in
the mycelium the cells are older. The old cells contain more and more
vacuoles (see Figure 4.2) and less cytoplasm and these eventually die and
parts of the mycelia are then broken down by autolysis mechanisms.
The production of secondary metabolites, e.g. antibiotics, which are
common products in molds, is not evenly distributed amongst the cells
and they are believed to be formed in the newer parts of the cell, not
however in the tip but in cells immediately behind it.
Yeasts belong to the fungi but the production microorganisms rarely
form a mycelium. This includes the common production microorganism:
Saccharomyces cerevisiae. Some forms, like Candida, are however able to
switch between the general growth form of “budding” and hyphal
growth.
A yeast cell also starts the cell division by the formation of a bud but not
from a spore but from a viable cell. As this bud grows, the content of
the mother cell partly flows over to the daughter cell and cell division
follows. Eventually, the daughter is separated from the mother, leaving a
scar on the cell surface of the mother. The mother cell only has room for a
certain number of scars and can thus give birth to only a limited number
of new cells. Consequently, we can here also talk about a cell age. Figure 4.3: Budding yeast cell. Note the char-
acteristic scar from a previous budding at the
Despite the cell adaptation to different growth mechanisms, we generally bottom of the cell
consider that growth of all types of cells is exponential, i.e. the total mass
of cells is doubled in a characteristic time interval. This time interval
is of great practical importance and is called “the generation time” or
"doubling time". This generation time is proportional to “the growth
rate” as will be shown in the next part of this chapter.
A general conclusion about the growth of different cell types, is that the
higher growth forms, the eukaryotic cells, seems to have a limited life
span in submersed cultivation (production conditions), while bacteria
seem to be able to grow “forever”.

4.2 Storage of production cells

Before we consider growth in more detail, it is relevant and actually most


important to consider the pre-history of the inoculum, i.e. we want to
know, down to the very detail, - and control - how the starting cellular
material for has been stored and propagated.
40 4 Growth of Cells

Our starting point for gathering knowledge about a particular cell is the
information given when the cells are developed, purchased or licensed
from some particular source. There are large commercial providers of
cells of which one is the American Type Culture Collection or, ATCC
(http://www.atcc.org), which contains a large number and variety of
different organisms. The cells are described by “a code” and an annotation
which indicates the origin of the cell and its genetic make-up and a
purchase will also give access to the background as far at it is known.
This should be very carefully studied since it contains a lot of information
relevant for the process design. Here is stated e.g. whether the cell is
an auxotroph with respect to certain nutrients e.g. whether one or more
amino acids need to be added to the medium. For the cultivation of E. coli
the genotype will tell whether the cell is sensitive to recombination (recA,
Hfr) or to phages (F factor). The origin of the cell also shows whether
it has an E. coli K12, E. coli B or another ancestor background, and this
says e.g. something about the cell morphology and particulars of the
metabolism.
For recombinant protein production we will also provide a vector (plas-
mid), a circular DNA fragment, to be introduced into the cell. Some such
vectors are naturally present in particular cells and provide these cells
with specific features. Such vectors are commercially available and can
be the starting point for modifications including the addition of the gene
for the product in question. Typical information in vectors are the control
elements for their replication and information of antibiotics resistance
genes that may be used to ascertain the continued presence in the cell.
Here is a lot of information that needs to be taken into account before the
cultivation design can even be started!
The production cells are often stored in -80°C freezers, in what we call
“cell banks”. The “bank” is used to preserve some important features of
the production cells: their identity, their purity and their suitability for
production. Their handling is particularly important when genetically
unstable cells are used. We start the storage procedure by preserving
a set of “Master cells” where a “Master Cell Bank” forms the genetic
basis of the production cells. In the freezer, we also have the “Working
cell bank” (WCB) which is derived from one vial of the “Master cell
bank”, and the WCB should be large enough to satisfy the needs of each
production run planned for “eternal times”. Before preservation, all cells
will be suspended in a specific medium in order to sustain their integrity
during storage. This is also added to protect the cell during the freezing
or freeze-drying and thawing procedures. The formula often contains
a solution based on glycerol. The characteristics of the stored cells are
carefully studied with respect to many parameters to ensure that the cells
maintain the qualities for which they were designed. One such parameter
is the growth pattern.
For the preparation of the Cell bank, we need a certain number of cells to
put into each vial and we need these cells to be vital, i.e. metabolically
active and growing. If the bank consists of cells and not vegetative spores,
Figure 4.4: First steps in inoculum prepara- this is usually done by cultivation of the cells in shake flasks in a medium
tion: agar plate and shake flasks. The smear- that is favourable for growth. The medium can vary depending on cell and
ing on the agar plate is done so that single
company preferences. In the handling, it is very important to reduce the
colonies are appearing hereby ensuring a uni-
form genetic background when the cell bank is number of generations since the number of possible mutations increases
established with each additional generation. This is one of the reasons to why we
4.3 Inoculum preparation 41

keep a set of cells from this original batch in the freezer which is large
enough for all future production, i.e. in order to be able to guarantee that
the genetic background is the same. The cells are allowed to grow into
the so-called "exponential-" or "logarithmic phase" (see below) and then
harvested. After the cell mass has been estimated, a selected amount of
cells is suspended in the appropriate solution for freezing.
If the production cell is a fungus, which generally forms a mycelium, the
inoculum will be in the form of vegetative spores. These can be obtained by
cultivation of mycelium on e.g. sterile rice, which is inoculated after being
carefully soaked in a small quantity of sterile water. When the mycelium is
not submersed in water, aerial hyphae, and later conidiophores, develop.
From these, the spores (conidia) are generated, and these can be dispersed
and harvested in an appropriate aqueous solution. This solution is
distributed into vials after counting. Cultures, particularly of spores, can
often be freeze-dried instead of frozen.

4.3 Inoculum preparation

Thawing of a vial from the Working Cell Bank starts the propagation
of the cells. Since the cell number (or number of spores) in the vial is
known, this information is used in the subsequent propagation. If the
vial contains cells and not spores, the vial is often used to inoculate an
agar plate, i.e. the cells are initially allowed to grow on a solid (agar)
support.
Growth on agar is a very specific growth situation where the cells
will grow in colonies. Access to oxygen, and nutrients, will depend on
where in the colony the cell is placed. This means that the cells are in a
physiologically very different state on an agar plate than when they are
dispersed in a liquid medium.
The agar plate is harvested and the cells are further propagated in shake
flasks, which means that growth takes place for the first time by cells
submersed in a medium. The media in the shake flasks can be of various
compositions. Since it is very important that the cells start to grow, a
complex medium may be used so that the cells are given the very best
opportunities for growth and thus need not themselves synthesise a lot
of the components that are instead provided in the medium. On the other
hand, we do not want the cells to go through many generations under
conditions other than those in the production bioreactor. This may lead
to mutations that favour growth in this particular environment since the Figure 4.5: Aspergillus conidiophore and
spores. Spores are collected by harvest pro-
cells will develop a physiological state best fit for survival. If the latter is
cedures, stored and used for propagation of
a risk, it can be overcome by the use of production media very early on new cells
in the propagation.
The methods of propagation used industrially are often empirical in the
sense that the procedure has been used many times and has been found
to be sufficient for the needs of a particular cell and process. It should
however be noted that if a new cell and product are to be developed, one
cannot automatically rely on data from other processes. If the propagation
is based on spores these are generally inoculated directly into a liquid
medium. The cells are propagated in a step-wise fashion in larger and
42 4 Growth of Cells

larger volumes and finally into the production bioreactor. The total number and size of such steps needs to
be designed at each occasion and is depending on the available equipment. Why do we not inoculate the
production reactor directly? Well, it is expensive to use the production fermenter for a very long time at very
low cell densities early in the exponential propagation.

4.4 Growth patterns

When a microorganism has been inoculated into a medium in a bioreactor, we can start to analyse how the
cell propagates i.e. how it grows. Several measuring techniques for this purpose have been described in the
literature, and we shall return to these in the last part of this chapter. Let us for the time being merely say
that we determine the increase in the cell mass by weighing a sample of the cell suspension at regular time
intervals. This often, but not always (see later), results in a “standard growth curve” that is often shown in
elementary microbiology textbooks.

The phases of a growth curve

The growth curve has a specific time-dependent shape which shows that the growth passes through several
distinct phases. Such a curve and its phases are shown in the Figure 4.6.

Exponential Stationary
Lagphase phase phase

25

20

15
Mx(g)

10

0
0 5 10 15
Time(h)
Figure 4.6: “Standard growth curve” (in blue) for a batch cultivation process and its different phases marked by red lines

The first phase, i.e. the first four hours in the example curve shown in Figure 4.6, begins immediately after we
have inoculated the bioreactor. In this phase, the cells may initially not grow at all, grow only very slowly
largely depending on their prehistory, or only a very small fraction of the cells start growing. This adaptation
phase is called “the lag phase”. It should perhaps be pointed out that all the cells do not divide at the same
time. Only under some very specific conditions can the cells be steered to multiply all at the same time and
this phenomenon, called “synchronised growth”, is only rarely seen, or supported, in production processes.
One reason why the cells do not grow faster in the lag phase is that they have to adapt to the growth
conditions in the new bioreactor where the micro-environment most likely is different to the previous one.
The cells may also suffer from the transfer from one reactor to another meaning that we should take care that
this is carried out as smoothly as possible. This cellular adaptation process involves growth control functions
4.5 The rate of growth 43

that can be anything from enzyme level adaptations (rapid) to the full induction of a set of new operons
(hours). It is here that knowledge of the inoculum history becomes relevant – the greater the difference
between the inoculum conditions and those in the production bioreactor, the longer is the lag phase. In
contrast, if there are no difference there might be no lag phase at all.
A common situation in which conditions are radically altered between reactors is the situation when cells
are initially propagated on an agar plate on one medium (solid phase growth) and are then transferred to
a bioreactor using another medium and where they will grow under submersed conditions. For example,
going from solid agar growth on a complex medium containing amino acids as nitrogen source and glycerol
as carbon source to a defined liquid minimal medium with glucose and ammonium as carbon and nitrogen
sources, respectively, will certainly require some adaptation and the lag phase may be long.
After becoming used to the new conditions, the cells will start to divide as fast as they possibly can under
the present conditions. The actual rate depends on many things as we shall see, but each cell type will
certainly grow at a different rate even if conditions are very favourable. When cells grow exponentially at
the maximum rate, they are in the “exponential growth phase” (sometimes also called "log-" or "logarithmic
phase"). Sometimes we talk here also about the concept of “balanced growth” by which is meant that all the
cellular material (e.g. ribosomes) is generated at the same rate as the growth rate and that the cell weight is
constantly the same. As we will see later, this is an idealised concept which may not prevail for very long, but
will depend on how we design the cultivation.
At some point, the cells will run out of one of the elements needed or encounter an environmental limitation.
The abrupt cessation of growth shown in the growth curve in the above figure indicates that this element
is carbon and the background to this statement will be explained later on. Actually, most carbon sources
that are actively taken up by the cell, will lead to this immediate cessation of growth and this is possibly the
only case when the growth curve has this shape. Since carbon is often added in the form of a sugar, often a
monosaccharide and in a known amount, we may be able to postulate that it is the lack of this sugar that
forces the cells into the next phase called “the stationary phase”. In this phase, no cell division is observed,
even though the cells may very well be metabolically active. Later, we shall see how the exhaustion of other
substrates leads to the development of other growth curves where it is difficult to discern a stationary phase
as it is defined above.
Some cells will, in the long term, react to the lack of a substrate element (or environmental condition) by
dying, even if many cells may take a considerable time in doing so. For organisms commonly subjected to
variations in their access to nutrients – and many of them are – the viability can remain for a very long time.
When they do begin to die, this constitutes the final phase that may be distinguished in a “growth” curve
and is called the “the death phase”. Here cells may dissolve and vanish from the medium due to cell lysis,
and in this case the growth curve declines. In a production context, this phase is not very interesting since
we do not wish to steer the cultivation this far. For animal cells, which in the nature are not developed for
submersed growth, death by apoptosis or necrosis may occur already from the time of the inoculation and
we will rarely see the shape of “the standard exponential growth curve”.

4.5 The rate of growth

We are always interested in a good process economy, regardless of the product, and this means that we need
to know how rapidly we can accumulate a given amount of cells since a long production time means a high
cost. This is equal to an interest in the growth rate, which can be estimated from the slope of the growth curve
in the above figure. However, since this curve is exponential, the slope is not easily determined. A trick that
can be used in this situation is to instead plot the logarithmic value of the cell mass (Mx ) against the linear
time, in a so-called semi logarithmic curve, or we use a logarithmic scale on the y-axis, which is easily done
in most software programs, the former displayed growth curve may take the form shown in Figure 4.7.
The use of the logarithmic scale shows clearly when and how cells multiply or die in an exponential fashion,
since such phases appear as straight lines with particular slopes. In this new graph it is very easy to distinguish
44 4 Growth of Cells

Exponential Stationary
Lagphase phase phase

100

10
Ln(Mx [g])

1
0 5 10 15

0,1
Time(h)
Figure 4.7: Semi logarithmic plot of the cell mass (Mx ) against time. ln(Mx )=f(t)

the different phases, and we see here that the lag phase is absent and we only have a log phase. From the slope
of the exponential phase we can thus calculate the generation time, i.e. how fast the cell mass is doubled. The
generation time is denoted tg . It should now be obvious that the specific growth rate (or the generation time)
is rarely constant during the whole of the growth curve, but that the value is different in different phases.
There is however always a maximum growth rate and that is constant as long as the present conditions are the
same.
Another way of looking at growth is to consider growth as a chemical reaction, where:

𝑠𝑢𝑏𝑠𝑡𝑟 𝑎𝑡𝑒 + 𝑐𝑒 𝑙𝑙𝑠 → 𝑛𝑒𝑤 𝑐𝑒 𝑙𝑙𝑠 + 𝑒𝑛𝑒𝑟 𝑔 𝑦 + 𝐻2 𝑂 + 𝐶𝑂 2

The rate by which this happens, i.e. the slope of the curve describing the concentration of biomass against
time and this we call the volumetric growth rate (unit: g/L,h). This rate we can consequently express at each
point of the curve and (assuming a constant volume) we do this by use of the differential:

𝑑𝑐 𝑥
(4.2)
𝑑𝑡

where “cx ” is the common notation used for the cell concentration [in g/L], “t” is the time [in h]. In other
words, the rate at which the cell concentration changes over time. The general mathematical expression
which describes this change of cell mass with time is described by a differential equation which implies that
the cell mass increases exponentially:

Definition 4.5.1 Exponential cell growth:


𝑑𝑐 𝑥
= 𝑘 ∗ 𝑐𝑥 (4.3)
𝑑𝑡

where “k” is the rate constant. From this expression, it is evident that the rate of change is proportional to a
rate constant multiplied by the cell mass at that specific time point. We need cells to produce new cells! The
rate constant, k, is what indicates how fast this takes place and this constant can vary during the cultivation.
4.5 The rate of growth 45

In the cultivation of animal cells, where we will not accumulate cells to very high densities and where the
cells are relatively large, it is more common to count the number of cells. This does not change the kinetics of
growth and the equations so far described, but the expression “cell concentration" or "cx ” is replaced by the
“cell number" or "N”.
The rate constant for growth is very important and is therefore given a unique symbol, the Greek letter, 𝜇
[mju] which we use instead of “k”. It is called “the biomass-specific growth rate” and it has the unit [g of
cells/g of cells, hour]. It is called the “specific” growth rate since it is the cell mass that is produced with
respect to a specific amount of cells. In practice, we use the unit [h-1 ].
In order to be able to calculate the specific growth rate, we can integrate the above equation between two
points in time, of which we have an interest, i.e. where we have accumulated a certain known amount of cells
and where we know that the rate is constant. To understand over which intervals we have such a constant
growth rate, we should first make a plot of the growth. This will allow us to treat the specific growth rate as a
constant during integration and results in a simple equation. We substitute the constant “k” with 𝜇 in the
equation above and insert the values relating to the time window in which we are interested and this gives us
Equation 4.4.

∫𝑡 ∫𝑐 𝑥
1
𝜇 𝑑𝑡 = 𝑑𝑐 𝑥 (4.4)
𝑐𝑥
𝑡0 𝑐 𝑥0

Solving the integral:


𝜇(𝑡 − 𝑡0 ) = 𝑙𝑛(𝑐 𝑥 ) − 𝑙𝑛(𝑐 𝑥0 ) (4.5)
Rearrangement leads to an expression to calculate the specific growth rate between time t0 and t:

𝑙𝑛( 𝑐𝑐𝑥𝑥 )
𝜇= 0
(4.6)
𝑡 − 𝑡0

The doubling time (or generation time) is the time it takes for a cell culture to double in mass (e.g. cx0 = 1, cx
= 2). This can also be expressed by Equation 4.6 by some rearrangement, resulting in Equation 4.7.

Definition 4.5.2 Doubling time


𝑙𝑛(2)
𝑡𝑑 = (4.7)
𝜇

Equation 4.7 is an efficient way of calculating the specific growth rate if the generation time is known (and
vice versa). Equation 4.6 can of course also be further rearranged to give a suitable expression for the cell
concentration (cx ) at constant volume, (or mass (Mx ) if the volume is not constant) at any given moment
(Equation 4.9).

Definition 4.5.3 Cell concentrationa at any given moment


𝑐𝑥
𝑒 𝜇(𝑡−𝑡0 ) = (4.8)
𝑐 𝑥0

𝑐 𝑥 = 𝑐 𝑥0 ∗ 𝑒 𝜇(𝑡−𝑡0 ) (4.9)

a Cell mass ( 𝑀 𝑥 & 𝑀 𝑥0 ) can also be used


46 4 Growth of Cells

4.6 Influence on the growth curve of dead and/or non-viable cells

It should be noted that the growth rate described above is the net specific growth rate since a number of cells
may have died and have either been dissolved (and are thus absent from a cell count), or are still intact but do
not grow but are still counted. The net rate, that is the one we measure, may thus be incorrect and the 𝜇 of the
growing cells may in reality be higher. The mass balance for viable biomass needs to be adjusted for this and
instead we get Equation 4.10. The death of cells is never an easy factor to determine since we obviously cannot
count dead cells if they have disappeared (lysed) and the detection of dead but intact cells needs some type
of viability staining. While such staining procedures are readily available (see microbiology text books), if we
assume that dead cells stay intact, the mass balance for total biomass (xtot ) can be written as Equation 4.11.
If we were to have cells in the cultivation that are dying or we have a large inactive population (with respect
to growth), we might easily get a large error in our growth rate calculations and we may even experience that
cells do not grow according to our concept of exponential growth.

Definition 4.6.1 Mass balance for total biomass a

𝑑𝑐 𝑥𝑣
= 𝜇 ∗ 𝑐 𝑥𝑣 − 𝑘 𝑑 ∗ 𝑐 𝑥𝑣 (4.10)
𝑑𝑡

𝑑𝑐 𝑥𝑡𝑜𝑡 𝑑𝑐 𝑥𝑣 𝑑𝑐 𝑥 𝑑
= + = (𝜇 ∗ 𝑐 𝑥𝑣 − 𝑘 𝑑 ∗ 𝑐 𝑥𝑣 ) + (𝑘 𝑑 ∗ 𝑐 𝑥𝑣 ) = 𝜇 ∗ 𝑐 𝑥𝑣 (4.11)
𝑑𝑡 𝑑𝑡 𝑑𝑡
ak
𝑑 = death rate constant
c𝑥 𝑣 = viable biomass concentration
c𝑥 𝑑 = non-viable biomass concentration

From the growth curves in Figure 4.8 we might think that we have a lag phase, which would be an erroneous
conclusion due to presence of either “inactive” or “dead” but measurable cells. This situation is even more
pronounced if cells are actually disappearing, as can be seen in the right graph. Only an activity analysis
will help to elucidate the true situation. This is of course very important information, since it means that
something has to be done about the death rate (or the reason for inactivation) rather than the growth rate per
se.

Mx Mx
measured
measured

Mx
inactive
Mx active cells
Mx,i
cells
Mx,a dying
active cells Mx,a Mx,d
cells

Figure 4.8: Effects on the growth curve of inactive or dying cells. Left: one part of the population is growing at the maximum rate (Mx,a ) but a
constant part of the population is not participating in the production of new cells (Mx,i ). The latter is an example of VBNC cells (viable but non
culturable). Right: one part of the population is growing at the maximum rate (Mx,a ) but another part of the population is dying and disappearing
(Mx,d ). This pattern is very typical for growth in animal cell cultivation. Mx refers to cell mass (g).
4.7 Factors affecting the maximum growth rate 47

How may we get inactive cells? Well, cells might be damaged by conditions preceding the inoculation to
the bioreactor, i.e. from those prevailing during inoculum preparation. That cells are negatively affected
in this process is not surprising or unexpected, often conditions during e.g. shake flask cultivation can be
harmful since the pH is not controlled, the temperature may fluctuate, oxygen may be limiting and we may
overgrow the cells so that particular nutrients are exhausted. The transfer of the cells from one reactor to
another may also lead to effects on the cells, particularly if the transfer times are long and harsh shearing
conditions prevail. If the cells recover during the final step, this effect will be diminished with time.
Another situation refer to the definition of “living” and “dead”. The concept of “living” is often taken as
being synonymous with “dividing”. However, in some cultivations there may be a considerable amount of
cells that are viable but non-culturable (VBNC) i.e. cells that might have some metabolic activity but have lost
their ability to divide “for ever” and this may also lead us to think that the cells are not growing as fast as we
wish. As the cultivation continues the influence of this initial, and maybe large, proportion of dead cells
will of course diminish and the true growth rate will appear. In any case, this leads to deviations from the
standard growth pattern. It should be noticed that VBNC cells might still produce product.

4.7 Factors affecting the maximum growth rate

The maximum rate at which the cells grow on a chosen medium and environmental conditions (pH, temp.
etc.) is first and foremost dependent on the genetic background of the cell, so the first important factor
determining the maximum rate is the characteristics of the chosen cell. In general, the specific growth rate is
lower the higher the cell lies on the evolutionary scale, which means that 𝜇 for the common production cells
decreases in the following manner:

𝐸.𝑐𝑜𝑙𝑖 > 𝑌𝑒 𝑎𝑠𝑡 > 𝑀𝑜𝑙𝑑 > 𝐴𝑛𝑖𝑚𝑎𝑙 𝑐𝑒 𝑙𝑙𝑠

The decrease is not linear since the rate of growth of animal cells is very much lower than that of commonly
used heterotrophic microorganisms which have both a high affinity for the substrate and a high maximum
growth rate. The reason is that microorganisms have been developed as single cells and in competition with
a great number of other species in their natural niche and over many millions of years and where the law of
“survival of the fittest” was applied. This is of course not the case for cells grown in the body of an animal
which we now only recently are trying to cultivate in a submersed fashion.
ln (specific growth rate [h-1])
Specific growth rate [h-1]

Temperature [°C] 1000/K

Figure 4.9: Effect of temperature on the specific growth rate of E.coli cells grown in rich medium. Left: The maximum growth rate possible is here
close to 2 h-1 and is gained in an interval between 39 and 42 °C. The corresponding value on minimal medium will be closer to one. Right: The
corresponding Arrhenius graph shows with the dotted line shows the interval where the Arrhenius equation can be used to postulate the rate. The red
arrows show the cultivation interval where processes normally operate
48 4 Growth of Cells

The second most important factor affecting the maximum growth rate is the temperature. In the curve for E.
coli in Figure 4.9 (left part) we can clearly see a maximum rate around 40°C and that at both higher and lower
temperatures, the value of the growth rate drastically drops. The reduction in growth rate is however much
more pronounced when approaching high temperatures. The cells grow in this experiment in a rich medium
i.e. containing amino acids wherefore the maximum growth rate is close to 2 h-1 which will be approximately
the double rate compared to a minimal medium at the same temperature.

Definition 4.7.1 Growth rate dependence of temperature

𝜇𝑚𝑎𝑥 = 𝐴 ∗ 𝑒 Δ𝐸/𝑅𝑇 (4.12)

Δ𝐸 1
𝑙𝑛(𝜇𝑚𝑎𝑥 ) = − ∗ + 𝑙𝑛(𝐴) (4.13)
𝑅 𝑇

The dependence of the specific growth rate of the temperature is often approximated by an expression
(Equation 4.12) similar to the Arrhenius plot for chemical reactions, where A is an organism-dependent
constant, ΔE is the activation energy of the reaction, R is the universal gas constant (J/mol·K) and T is the
temperature (K). Values of these constants for different types of cells can be found in the literature. However,
if the growth rate at different temperatures is known, the logarithmic value of the equation will lead to an
expression (Equation 4.13) that can be plotted as a straight line in a graph of ln 𝜇𝑚𝑎𝑥 =f(1/T).
The slope gives then the activation energy. However, from the Figure 4.9 (right part) we can see that the
linearity is kept only for a limited part of the temperature interval. The reason is that the temperature also
influences many other reactions in the cell that tend to influence the growth rate. For E. coli, this interval
is between 20 and 37 °C where other growth interfering cellular reactions are often unaffected by the
temperature and reactions are modified by the enzyme activity only. During all temperature changes the
cell membrane composition of phospholipids and fatty acids will change to keep the desired fluidity. In
practice, cultivations are most often performed at 30-37°C and this may also have to do with the possibilities
for efficient control of the temperature in the presently used bioreactors.
The third important factor that affects the maximum growth rate for a given cell at a given temperature is the
medium composition as stated before. E. coli growing on a minimal medium must itself make a number
of building blocks for cell growth that might be provided other media. This involves the making of e.g. all
amino acids, fatty acids and vitamins that are needed in the building of a new cell, and this reduces the
maximum specific growth rate. When E. coli has access to all these building blocks at a temperature of 37°C
the doubling time can be as short as 15-20 minutes and the specific growth rate faster than 2 h-1 , - a rate which
is unparalleled in other organisms.
The media used for growth are also characterised by a specific pH and the acidity of a solution is another
important factor which will influence the maximum growth rate in a particular medium. It should be noted
that even though the pH of the medium changes, the intracellular pH could be kept fairly constant due to cell
homeostasis. This may however be done at a certain cost to the cell since energy may have to be used. Different
organisms may grow under very different pH intervals. For E.coli the pH range is 5-8, yeast grow well at pH
6, but some microorganisms are e.g. very acid tolerant. This is particularly true of cells that accumulate acids
or have developed in nature under highly acidic conditions. The effect of pH on the maximum growth rate of
E. coli is shown in Figure 4.10 at different temperatures.
Coupled to the medium composition is the concentration of medium components. Both high and low
concentrations may reduce the growth rate. A high concentration of a medium component might inhibit
growth, and at a low concentration, the maximum rate for that medium composition cannot be upheld.
The specific case of production of a recombinant protein in a host cell, i.e. the fourth product category
described in chapter 1, may also influence the growth rate. Here, foreign DNA has been introduced into the
cell – a situation for which it certainly has not been developed. The replication of both the chromosome and
4.8 The time-dependent uptake of sugars 49

Specific Growth Rate [h-1]


0.95

0.9

0.85

0.8

0.75

0.7

0.65
4.5 5 5.5 6 6.5 7 7.5 8 8.5
pH
Figure 4.10: Effect of pH on the maximum specific growth rate, 𝜇 [h-1 ] of E. coli cultivation in minimal medium. The three curves show the effect at
different temperature. Red: 37°C, blue: 39°C and green: 41°C

the plasmid vector carrying the foreign DNA may in many cases be a metabolic burden and lead to a slower
growth rate even without the induction of the product. This may also be the reason why some plasmids
cannot be used at all, since the cells will not grow due e.g. to the energy cost of replication.

4.8 The time-dependent uptake of sugars

When the cells grow there is naturally also a simultaneous consumption of nutrients. Let us therefore now
also consider the consumption rate also of the nutrients as a function of time. We will start by looking at the
consumption of the carbon source which is often a monosaccharide and very often glucose.
When a cell takes up carbon it is mainly used for production of new cells and to produce energy to support
the growth as we have noted before. It is really fair to say that a main obstacle for a cell is its own reproduction.
We will now call this our “model” for cell growth and we can illustrate this model by Figure 4.11. The
biomass-specific substrate consumption rate has also its own symbol (as has the specific growth rate i.e. 𝜇)
and is given the symbol “qs ” with the unit: [g substrate/g cells·h]. A graphical representation of the model
can be seen in Figure 4.11.

Figure 4.11: Model of cell metabolism of sugar which leads on one hand to growth and new cells and on the other hand to energy generation which is
used to build the new cells. The letter “q” is used to describe the specific rate of consumption of carbon for the various purposes

In Figure 4.11 we consider that the flux for intake of sugar (or the specific consumption rate) to the cell, qs , is
divided into two fluxes inside the cell: one where the carbon ends up in the new cells and the second where
the carbon ends up as carbon dioxide as a result of the energy generation.
50 4 Growth of Cells

Mx (gx)

Ms (gs)
Figure 4.12: The growth of cells (Mx ; blue) and the simultaneous consumption of glucose (Ms ; red). No lag phase is present. Note that when the cells
enter the stationary phase this is simultaneous to exhaust of the glucose

If we would now also measure the glucose concentration at the same time as we measure the cell growth and
if we plot the values together with the cell accumulation data, which we looked at before, this might look
something like Figure 4.12.
Initially, the decrease in glucose concentration is very low because there are very few cells present to consume
the initial sugar, i.e. the inoculum is low. This should not be confused with a lag phase, which is not present
in this cultivation. This means that the few cells that are present grow at their maximum rate and thus take
up the sugar at the maximum rate. Again, a semi-logarithmic plot will give a straight line for the growth as a
function of time if cells grow exponentially. From the sugar analysis, we may now also verify that the cells
have entered the stationary phase due to exhaust of the sugar since the sugar concentration (S) will be zero!
We may of course also want to calculate the specific consumption rate, in analogy with the specific growth
rate, which is given by the slope of the curve at any time. We may use the same type of expression as in the
case of growth, where the rate constant is replaced by the symbol, qs . The negative sign is of course due to
that the substrate is consumed. When the equation is solved, we will see that we still get a positive sign for
the specific consumption since we go from high to low concentration. This differential equation is not as easy
to solve as the one for growth. This is due to the fact that both the cell concentration (cx ) and the substrate
concentration (cs ) are a function of time (t).

Definition 4.8.1 Rate of substrate consumption (assuming constant volume)

𝑑𝑐 𝑠
= −𝑘 ∗ 𝑐 𝑥 = −𝑞 𝑠 ∗ 𝑐 𝑥 (4.14)
𝑑𝑡

∫𝑐 𝑠 ∫𝑡
𝑑𝑠 = −𝑞 𝑠 𝑐 𝑥 ∗ 𝑑𝑡 (4.15)
𝑐 𝑠0 𝑡0

A trick for approximating qs is to consider the cell mass as being constant over a small interval and solve the
integral in the following way: 𝑞 𝑠 = −(Δ𝑐 𝑠 /Δ𝑡)/𝑐 𝑥 . The integral can only be solved in this way over a minor
time interval in particular if there are major changes in the variables. Which value for “cx ” should then be
used solving the equation? We can estimate the specific substrate consumption by using either of the known
cx -values (i.e. cx at time t=0 or at t=t). Please observe that the value in between these two points we know
nothing about (unless we estimate it from exponential growth).
4.9 The substrate uptake rate as a function of substrate concentration 51

qs [gs/gx,h]
Ms [gs]
Mx [gx]
μ [h-1]

Figure 4.13: Figure showing the specific consumption rate of sugar (green) and specific production rate of cells (black). Please observe the different
scales in particular with respect to the rates. The substrate uptake rate is always higher than the growth rate since more sugar is needed due to the
carbon dioxide production!

In Figure 4.13, the values of the two specific rates are plotted together with the cell and substrate data. We can
here indeed see that the values both for the specific growth rate and the specific glucose consumption rate
are constant over the whole interval since there is no lag phase. This corresponds to, and is the very meaning
of, the definition of the “exponential growth phase” i.e. it is here that we find the constant maximum specific
growth rate and a continuous exponential increase of the total cell mass. In the stationary phase, the specific
growth rate will consequently be zero.
The curves in the graph might suggest that there is a correlation between the cell mass increase and the
glucose consumption – the curves look like mirror images of each other. And indeed there is! This correlation
is a most important parameter in cultivation technology and indicates how many cells that can be obtained
from a given amount of substrate. This “efficiency factor” is called “the yield” of cells per substrate and is
denoted “Yx/s ” (e.g. in gx /gs ).

Definition 4.8.2 Yield of cells per substrate


𝜇
𝑌𝑥/𝑠 = (4.16)
𝑞𝑠

In this case, the yield is defined as the quotient between the specific growth rate and the specific sugar uptake
rate as shown above. This concept will be discussed into greater detail in a later chapter

4.9 The substrate uptake rate as a function of substrate concentration

We have several times pointed out that not only the choice of compounds of the medium but also their
concentration is important. When we design a process we are therefore interested in knowing how the
substrate concentration affects the specific growth rate or more clearly, we need a model that shows how the
specific substrate consumption rate depends on the substrate concentration.
What could be parts of such a model i.e. what is controlling the rate? The answer is that this is almost solely
controlled by the characteristics of the uptake systems for the nutrients that are located in the cell membrane.
From the lessons in microbiology or biochemistry we have learnt that there are two principal types of
transporters: (1) those controlled by diffusion and thereby of the concentration gradient of the substrate
between the interior and the exterior of the cell and (2) those that rely on active transport by carriers, often
52 4 Growth of Cells

against a concentration gradient. As we have also remarked here, several different systems may be active for
the transport and this means that a combination of these may be working at the same time. If a substrate
is taken up by diffusion (case 1 above) the uptake rate can be described as a simple linear function of the
concentration and the rate (v) can be expressed by: 𝑣 = 𝑘 ∗ [Δ𝑐 𝑠 ] over the membrane area, where k is the
permeability coefficient and a function of the diffusivity.

Figure 4.14: Rate of substrate transport against substrate concentration.

This curve (Figure 4.14) shows principally that the higher the substrate concentration the higher is the uptake
rate. This model might be relevant for describing for example the uptake of inorganic ions such as nitrogen,
phosphorous and sulphate. Note that this does not lead to the establishment of a maximum rate - here the
rate increases probably up to a point when the substance becomes toxic which is setting the limit.

4.10 The Monod equation

In the case of a carrier-mediated solute transport over the phospholipid membrane, the mechanism is radically
different and consequently also the rate expression. This situation is generally described by a model for
the so-called “saturation kinetics” which principally means that we say that there is a maximum rate to
be reached by which the substrate can be transferred. This model is very similar to the Michaelis-Menten
expression used in enzyme kinetics. The French scientist Jacques Monod first suggested the mathematical
expression for cultivation concepts (Equation 4.17), where, “cs ” is the variable substrate concentration (g/L)
and “𝐾 𝑠 “ is the substrate concentration (g/L) at which the rate is half of the maximum ( 𝑞 𝑠𝑚𝑎𝑥 or 𝜇𝑚𝑎𝑥 ) also
called the saturation constant.

Definition 4.10.1 Monod’s equation for growth


𝑐𝑠
𝜇 = 𝜇𝑚𝑎𝑥 (4.17)
𝐾𝑠 + 𝑐𝑠

Monod’s equation adapted for substrate consumption


𝑐s
𝑞 𝑠 = 𝑞 𝑠𝑚𝑎𝑥 (4.18)
𝐾𝑠 + 𝑐s

Equation 4.17 shows the dependence of the growth rate on the concentration of substrate which is what Monod
originally measured. This relationship is in cultivation technology therefore called the Monod equation or
Monod kinetics. The equation can equally well be written on the basis of the rate of substrate uptake which
4.10 The Monod equation 53

is shown as Equation 4.18. From this equation it is evident that only if “cs ” is “very high” (i.e 𝑐 𝑠 >> 𝐾 𝑠 ) will
the specific growth rate be at its maximum otherwise it will just be a quotient of this.
From the Figure 4.15, we see that when we operate at high substrate concentrations, the growth rate is not
influenced by changes in this parameter. We call this “zero order kinetics” i.e. when the rate constant is
independent of the concentration of the substrate and 𝜇 = constant = 𝜇𝑚𝑎𝑥 . This is the maximum specific
growth rate.

qs [gs/gx,h]

μ [h-1]
Figure 4.15: Monod plot for growth of E. coli on glucose. The plot is made both for the growth rate (red) and the specific substrate consumption (blue)
described by the Monod kinetic expression above.

If we move along the curves to lower concentrations, there is however a strong dependence on the substrate
concentration and the actual rate which is given by the equation. Note in particular two things: the dramatic
decrease in rate that follows below a very distinct concentration (which is of about 50 mg/L for E. coli), but
also the very low concentrations that still give very high rates – we are in the milligram/L area when we still
have half the maximum growth rate!
When cells are cultivated, the nutrient that is often most important is the carbon source and this is often
a monosaccharide which can easily be translocated over the membrane. Each carbon source has however
different values with respect to the constants in the Monod expression. The uptake mechanism is different for
different sugars and the cell “likes” some sources better than others and the “most liked” will have lower
values of the saturation constant and sometimes also higher values of the maximum uptake rate. Figure 4.16
below shows E.coli growth on two different carbon sources, glucose and glycerol with different 𝐾 𝑠 values and
the same maximum rate.
qs [gs/gx,h]

Figure 4.16: Monod plot for the growth of E. coli on glycerol (blue) and on on glucose (red). The maximum substrate uptake rate is about the same but
the saturation constants radically differ (110 and 5 mg/L, respectively)

In this graph it is evident that the saturation constants for these two substrates are very different. The
interpretation is that E.coli “likes” glucose better, or maybe in more stringent mechanistic terms, that the
uptake system for glucose has a higher affinity for its substrate than the glycerol system has for glycerol. The
54 4 Growth of Cells

maximum rate is interpreted as the situation at which each transporter is saturated with substrate and this
rate is almost the same for both substrates. The saturation constant, or the Ks -value, has however a strong
influence on the shape of the curve and over most of the concentration interval of the graph the uptake rate
of glycerol is lower than the uptake rate of glucose. This means that a higher glycerol concentration is needed
to reach the same rate of uptake as for glucose. It is generally so that there is a higher Ks -value for substrates
other than glucose. This is one of the reasons why glucose is the preferred substrate in many organisms.
The Monod curves for different organisms grown on the same carbon source and the same concentration
interval also differ. Figure 4.17 shows the relationship of two common production microorganisms, E. coli and
S. cerevisiae grown under standard conditions with glucose as carbon source. The saturation constants for
these two microorganisms are very different, 5 and 180 mg/L, respectively, and the maximum rates also differ.
This is evident from the graph where the yeast is not growing at half its maximum value at a concentration at
which E. coli is growing at its maximum rate. Under such conditions E.coli will outgrow the yeast if they are
inoculated at the same time and at the same inoculum size.
qs [gs/gx,h]

Figure 4.17: Monod plot for the growth of E. coli (blue) and S. cerevisiae (red) on glucose. Both the maximum growth rate and the saturation constant
differ (qsmax =1.6 and 0.9 and Ks =5 and 180 mg/L, respectively)

The difference between yeast and E.coli is due to different uptake mechanisms for glucose. While E.coli has
developed a high affinity system for the sugar uptake and can take up sugar through different systems, yeast
has to depend on facilitated diffusion which is generally less specific and requires a high concentration
gradient to result in a high uptake rate. However, even if this is diffusion-controlled it is carrier-mediated,
which leads to the establishment of the Monod model kinetics.
At the very high glucose concentrations to the right in the figure, the Monod model states that the substrate
uptake rate is constant which should also apply to the growth rate, provided that there is a constant
relationship between the two. This is not altogether true. The growth rate can in reality be reduced at very
high glucose concentrations due to many factors. Increasing concentrations of sugar leads to a lowering of
the water activity, which will eventually become too low for growth. High sugar concentrations also affect
the osmotic pressure and this may e.g. lead to stressful situations inside the cell. For E. coli cultivation, such
effects can be present already at a glucose concentration > 10 g/L.
The Ks -value will thus differ with different organisms and different substrates. Some examples of Ks -values
are given in Table 4.1.

Deviations from the standard growth pattern

The standard growth curve with lag, log and stationary phases actually represents only a very specific
case of growth. This curve describes growth on glucose as the carbon source and where this substrate was
added in an amount that it would eventually be the first to limit the growth.
4.11 Batch growth with limiting substrates other than the carbon source 55

Table 4.1: Examples of Ks -values for different microorganisms on different substrates

Organism Substrate K𝑠 (mg L−1 )


Escherichia coli Glucose 5
Escherichia coli Glycerol 110
Escherichia coli Lactose 20
Escherichia coli Phosphorus 2
Saccharomyces cerevisiae Glucose 180
Aspergillus oryzae Glucose 5
Penicillum chrysogenum Glucose 4

There are several reasons to why we may get a different growth pattern and the most common are a result
of:
I Use of another limiting substrate than the carbon source
I Use of mixtures of different carbon or nitrogen sources
I Cellular production of growth inhibitors such as protons (low pH), toxic compounds and overflow
metabolites
I Growth limitation by limited access to oxygen (feed limiting amounts of air)
I Poor affinity for the first substrate to limit the growth

4.11 Batch growth with limiting substrates other than the carbon source

What happens to the growth curve if we choose another of the Macro-elements, say nitrogen, or phosphorous
to become limiting for the growth? This is easily done during medium design where one of these components
is added in a low concentration while keeping the sugar in excess. Let us first consider the case when nitrogen
is the first element to be depleted during the cultivation.

Ammonium
cammonium [mg/L], cglucose*20 [g/L]

Protein
Cell mass
cx (gx/L)

Glucose
Carbohydrates

Time (h) Time (h)

Figure 4.18: The graph is showing the development of selected variables during nitrogen depletion. Left: cell mass production (blue) glucose
consumption (red) and nitrogen consumption (green). Right: the accumulation of protein (green) and carbohydrates (blue). The dotted line shows the
time of nitrogen depletion

From the growth curve in Figure 4.18, it is obvious that we cannot strictly talk about a stationary phase as
in the case of glucose depletion. When nitrogen, here in the form of ammonium, is finished, the cell mass
actually continues to very slightly increase, but in a linear fashion. After the ammonium is depleted glucose
is still available and is indeed taken up by the cell as shown by glucose analysis. As in the case where glucose
was the limiting source, however, there is an immediate deviation from the exponential growth pattern when
the nitrogen is depleted since there is a diminished possibility of building DNA and RNA which is obviously
very important for sustaining the growth and for which a lot of nitrogen is needed.
56 4 Growth of Cells

The continued increase in cell mass after nitrogen depletion is in this case is mainly a result of the cell
accumulation of carbon-based polymers i.e. the cells become more dense although they do not increase
in number and this would consequently only be seen when cell growth is measured by weight. Under
such circumstances, the cells search for alternative nitrogen sources and one solution can be to consume
cell-internal amino acid-rich substances e.g. from the degradation of ribosomes or proteins. This is an example
of a deviation from the concept of “balanced growth” since macromolecule production and growth are no
longer coupled. The carbohydrates that are during such cases accumulated takes place at the expense of the
cellular fraction of protein, which is clearly seen to the right in Figure 4.18.
If phosphate is depleted, the story is very different as can be seen in Figure 4.19 where the growth curve
continues to increase with time (although not exponentially) even after the phosphate concentration is zero.

Total carbohydrate, total protein [% cell mass]


6 80 100

cPhospahte [mg/L], cglucose*5 [g/L]


Phosphate Cell mass 70
5
80
60
Protein
4
50
60
cx (gx/L)

3 40

30 40
2
Glucose Carbohydrates
20
20
1
10

0 0 0
0 2 4 6 8 10 0 2 4 6 8 10
Time (h) Time (h)

Figure 4.19: The graph is showing the development of selected variables during phosphate depletion. Left: Cell growth (blue), glucose consumption
(red) and phosphate consumption (green). Right: The accumulation of protein (green) and carbohydrates (blue). The dotted line shows the time of
phosphorous depletion

Also in this case, there is an accumulation of carbohydrate containing polymers resulting from the uptake
of glucose as in the case of nitrogen depletion. No internal nitrogen source is consumed since the medium
contains nitrogen and the protein level stays thus constant. In this situation the cells are clearly continuing
to multiply, which indicates that there must be an internal source of phosphorous. All cells have a pool of
phosphate, often in the form of poly-Pi and this internal source is now being used to produce new cells. In
other words, the phosphorous available is diluted over more cells.
When nitrogen or phosphorous is depleted and the cells still have access to a lot of carbon this is taken up in
spite of the restricted possibilities for growth. Some cells types will start to produce the specific polymer
metabolites discussed above. The reason is often not known, but it can be a way for the cell to store carbon
for harder times. This phenomenon has in fact been used for the commercial production of a variety of
products.
Oxygen is often depleted during cultivation because it is difficult to supply to a bioreactor in the large
amounts needed. This is due to its low solubility in water and the fact that it is used at a rapid rate for the
respiration of many cells. In bioreactors, oxygen can thus often not be supplied in the amounts needed,
particularly if we wish to grow cells to high density at the maximum rate. The lack of oxygen will influence
the growth pattern and this is seen as a continuously decreasing growth rate, i.e. as a deviation from the
exponential growth pattern.
A possible lack of oxygen is most easily detected in a plot of ln(x) against time, which will start to deviate
from a straight line when all other nutrients are still in excess. The development of the process parameters
at oxygen depletion is very similar to the phosphorous-depletion curve. Only an analysis of the substrate
component levels will reveal which depletion is actually occurring.
4.12 Growth on mixed carbon or mixed nitrogen substrates 57

4.12 Growth on mixed carbon or mixed nitrogen substrates

What will happen if we mix two carbon sources in the medium? Will both be taken up simultaneously or is
there a preference for a particular carbon substrate? The answer is that it depends on the type of cell. More
specifically, it depends on the metabolic control of carbon uptake of the particular cell. Many microbial cells
have efficient systems for the uptake of carbon substrates to the extent that each cell may have two or even
more ways of taking up of a single substrate as we have stated before. The reason is, of course, its importance
for growth, and that there has been a continuous development of this capacity over millions of years, because
the availability of a specific nutrient is continuously changing in the environment. Nevertheless, the uptake is
quite unique for a specific cell and in many cases this is not particularly well known. If we are interested, we
have to find out for each cell and for each substrate.
Jacques Monod made many experiments with E. coli and a wide range of different mixtures of carbon sources,
and he found a growth pattern that clearly deviated from the standard pattern for growth on a single source
such as glucose.
Since the resulting curve showed two phases (see Figure 4.20), where each part was representative of the
consumption of one of the carbon sources, he called this pattern “diauxic” growth. Monod examined the
growth pattern for many combinations of sugars, and he found that if glucose was present together with
another monosaccharide it would always be preferred to other sugars. What could be the reason for this?

2.5

Xylose
2
Concentration [g/L]

Glucose
1.5

0.5

Cell mass
0
0 2 4 6 8 10
Cultivation Time [h]
Figure 4.20: Diauxic growth of E.coli on glucose and xylose. Note that xylose is not taken up before glucose is totally consumed and the cell has
rearranged its metabolism for the new carbon source

In E. coli, two control mechanisms are working to achieve this: “catabolite repression” and “inducer
exclusion”. They work in principle in this way: sugars are taken up through protein channels or by carriers
which we sometimes refer to as “permeases” and in E. coli these are found in the inner membrane. For
many substrates, like glucose, constitutive amounts of the glucose-specific permeases are always present.
The amount is small (why spend carbon and energy on something which may not be used?) but it can be
increased if the substrate is actually present in the medium. Sensors for this are can be found in the cell wall,
and they convey this information through chemical signalling to the inside of the cell. When the “catabolite”,
i.e. glucose, is taken up (by its own specific permease) and metabolised, the production of a central regulatory
molecule, cyclic AMP (cAMP), is radically reduced or even stopped.
This molecule is needed for the transcription of operons of permeases of other sugars so these can actually be
taken up. As long as there is an excess of glucose (the catabolite), it thus represses the uptake of the other
substrates, through lack of cAMP, since their uptake systems cannot be produced. Furthermore, other specific
metabolites that are present as a result of the high glucose uptake will effectively hinder the uptake of other
carbon sources by binding to the permeases of many of the other substrates is they happen to be present.
This is called “inducer exclusion” since the “inducer”, i.e. the new sugar, is not permitted to enter. All this
means that glucose is the preferred substrate and is always taken up first. After the supply of glucose is
58 4 Growth of Cells

exhausted, cAMP is again produced and the induction of new permeases can take place and, after a brief lag,
the next carbon substrate is metabolised. This lag phase takes usually 1-2 hours which is representative for
the genetically-based changes that are needed for increasing of protein levels.
Although we have described this mechanism for E. coli, catabolite repression is present in very many cell
types. In the case of Saccharomyces cerevisiae grown on a mixture of glucose and galactose, glucose will first be
taken up and when this supply is exhausted, the cell will induce its galactose carriers. If sucrose is added, the
growth follows the same pattern: after degradation by cell-wall-bound invertase glucose is first taken up and
fructose is thereafter consumed. It should here be observed that enzymes for disaccharide degradation are
not present in most organisms.
Nitrogen sources used in bioprocessing media are often nitrogen salts but they can also be in the form of
amino acids as indicated in the chapter on Media. The cell prefers to take up amino acids since they then
do not then have to waste building blocks and energy in making them. This leads to a high growth rate,
which in evolution must have been very desirable since rapid medium consumption would allow extinction
of a competitor. However, once the amino acids are finished, the cell itself must – if it is possible – start to
synthesise them. As in the case of growth on different carbon sources, this will take some time to achieve and
growth will meanwhile stop due to the need for enzyme induction.
This regulatory metabolism in E. coli in this case is called “stringent response” which was briefly mentioned in
a previous section. It is a general control mechanism which will affect a large number of cellular mechanisms
leading to cessation of growth. In this case, the mechanism is the following: the exhaustion of amino acids
leads to a lack of charged tRNA´s and to the production of a signal substance, guanosine tetraphosphate
(ppGpp) from GTP metabolism. This molecule, in turn, has the power to temporarily stall the ribosomes
by binding and a subsequent change of the translation initiation mechanism and further to hinder RNA
polymerase (RNAP) activity by ppGpp blocking of the sigma factor binding. This prevents protein synthesis
with the exception of proteins that during these conditions have a high affinity for RNAP which are controlled
by alternative sigma factors. As soon as the amino acids are again synthesised by the cell, growth continues.
As indicated before, an inoculum grown on amino acids in a rich medium followed by production on a
minimal salts medium will induce a stringent response which is probably undesirable since it will lead to an
operator-uncontrolled growth (and most likely production!) pattern.

4.13 Cellular production of inhibitory substances - overflow


metabolism

Many cell types are “peculiar” in that they take up carbon although they cannot fully metabolise it to carbon
dioxide and water. Just as we humans tend to eat an excess of food although we do not really need all of it,
leading to the storage of fat! Microorganisms also produce storage materials in various forms under such
conditions with lipids, glycogen or trehalose as common examples. Another common strategy of the cell that
takes up excessive sugar is to accumulate carbon in the form of metabolites which are excreted to the medium.
This means that the carbon is processed some distance down the catabolic route whereafter incomplete
oxidised metabolites are excreted. This is consequently called “overflow” metabolism and the metabolites
are called “overflow metabolites”. A maybe strange thing is that these metabolites are enriched as growth
proceeds even to a concentration that is eventually toxic to the cell. It should be noted that this cell-specific
accumulation of metabolites would occur even if the oxygen level is high and does not limit the growth. This
is a common metabolic strategy particularly in facultative organisms and many cells are subjected to this
although the metabolites differ.
Let us consider this overflow mechanism into more detail for some of our common production cells: E. coli,
Saccharomyces cerevisiae and animal cells.
When E. coli is subjected to a glucose concentration > approx. 30 mg/L or a growth rate (on minimal medium)
of approximately >0.3 h-1 , which is about 40% of the maximum growth rate, overflow metabolites will be
produced mostly in the form of acetic acid. The mechanistic reason to why this happens is not completely
4.13 Cellular production of inhibitory substances - overflow metabolism 59

clear and still the topic of scientific study. However, when glucose in the
medium is used up, many E. coli cells can indeed take up the acid again.
The acid production may in such a case be looked upon as a way to store
carbon in a format that can only be used by few organisms – i.e. a clear
competitive advantage in the nature.
E. coli forms acetate from two different metabolic routes, but the formation
through the conversion of acetyl-CoA is sometimes the more interesting
Figure 4.21: Simplified schedule of the over-
since it leads to the formation of ATP which is to the benefit of the cell flow metabolism in E. coli subjected to high
(Figure 4.21). This means that the cell loses carbon by the export of the glucose in the presence of oxygen. Glycolysis is
indicated by the multiple arrows between glu-
acid to the medium, but that in this process, still some energy is gained.
cose and acetyl-CoA which is also a precursor
for the TCA cycle.
The overflow metabolism of E. coli should not be confused with the
so-called “mixed acid fermentation” which takes place under very low
oxygen or anaerobic conditions, but which also leads to acetate formation
amongst other metabolites.
In a fungus such as Saccharomyces cerevisiae, excess glucose under strictly
aerobic conditions leads to ethanol production from part of the glucose.
The role here is most likely to regenerate NAD+ which allows a continu-
ation of the catabolism. The overflow metabolism starts when the rate
of glycolysis exceeds the rate of downstream metabolism from pyruvate
and onward, which in an over-simplified way is often attributed to a
bottleneck in the respiratory electron flow.
The origin of this phenomenon has been debated in the research literature
for many decades. However, when glucose is almost used up, S. cerevisiae
takes up ethanol again at the expense of ATP. Also here, the glucose
concentration for overflow is approximately 30 mg/L and the specific
growth rate 0.3 h-1 . The fact that yeast can tolerate higher ethanol
concentrations than any other organism naturally also gives it a growth Figure 4.22: Overflow metabolism in Saccha-
advantage. Also here, the ethanol formation should not be confused by romyces cerevisiae subjected to high glucose
concentrations in the presence of oxygen. The
the formation during anaerobiosis, the metabolism we use to produce dotted line shows the mitochondrial membrane
liqueurs, wine and beer. (mitochondrion below the line) over which
pyruvate has to be transported to generate
Animal cells can also demonstrate overflow metabolism. These cells are acetyl-CoA and access to the TCA cycle is
often grown on two carbon substrates, glucose and glutamine, from which permitted
both new cells and energy are acquired. The corresponding reactions
are called: glycolysis and glutaminolysis. In spite of available oxygen,
the metabolism leads to the accumulation of several byproducts like
lactate, alanine and ammonium ions. Both ammonium and lactate will
eventually inhibit cell growth.
Lactate is formed in substantial amounts from the uptake and metabolism
of glucose. The lactate promotes the regeneration of NAD+ and also
generates some ATP. The transfer of pyruvate to the mitochondrion is
also here a limiting step, meaning that only small amounts of pyruvate
enter the mitochondria for complete oxidation and energy generation.
The presence of the anaplerotic malic enzyme helps however in the
recycling of pyruvate from malate in the TCA cycle and this can be Figure 4.23: Overflow metabolism in animal
supplied by glutaminolyses. cells in the presence of oxygen. The dotted line
shows the position of the mitochondrion. The
block arrows show the overflow metabolites that
are excreted. Lactate is the product of glycolysis
and alanine the product of the glutaminolysis
60 4 Growth of Cells

Figure 4.24: Structured model of aerobic cell growth including overflow metabolism

The uptake of glutamine in glutaminolysis is followed by deamination to glutamate, the intracellular nitrogen
carrier. This leads to the production of ammonium which is excreted to the medium. If not used directly as an
amino acid, glutamate is metabolised by either of two routes, both leading to the formation of 𝛼 -ketoglutarate
and access to the citric acid cycle. The first route takes place through a transamination reaction which results
in alanine accumulation with pyruvate as a substrate. Alanine is thus an end product of glutaminolysis just
as lactate is the end product of glycolysis. In the alternative route, glutamate is deaminated which leads to
the production of a second mole of ammonium.
For all these cells types, the accumulation of overflow metabolites is toxic and the effect becomes more
pronounced as their concentration increases with consumption in prolonged cultivations to high cell density.
The cells will gradually reduce the growth rate as the toxic component accumulates. This is noted as a
continuous deviation from an exponential growth pattern with a continuous reduction in the specific growth
rate. This is again clearly seen in a semi-logarithmic growth plot, i.e. in the curve of ln(cx ) plotted against
linear time or cx plotted on a logarithmic axis. The pattern mimics to some extent the growth curve that
results from oxygen depletion and the different causes can thus easily be mistaken for each other if overflow
metabolites are not monitored.
The flux to overflow metabolism can now be incorporated into our metabolic model where we take the yeast
cell as the example (Figure 4.24). The lines in the figure are of course not drawn to scale since the production
of ethanol will reduce the amount of sugar that is available for new cells according to the discussion above.

4.14 Techniques to measure growth of cells

Many different techniques have been developed to estimate growth because this information is important for
process development and production. The final choice of technique will depend on the cell characteristics,
the cell density and the need for rapid on-line analysis data.
It is here appropriate to talk a little about cell morphology. Since we are examining the growth rate we may
also need to understand that the cell morphology is also a function of growth rate. The effects measured as
changes in “growth rate” may be a change in morphology such as a change in cell size and/or shape. The
cells may, in certain media and under certain conditions, aggregate to form clumps or become attached to
solid surfaces in the bioreactor. Depending on the analysis method, these phenomena may lead to more or
less erroneous results.
The techniques suggested for the estimation of growth build principally on an estimation of either the cell
mass or of the cell number. Some of the methods also give additional information about e.g. morphology and
viability, particularly if staining procedures are simultaneously introduced.
4.14 Techniques to measure growth of cells 61

The main principles are the following:

I Direct counting of cells (with and without the use of dyes)


I Scattering of light
I Weighing

Direct counting of cells

Cells that are not growing to particularly high densities and are rela-
tively large can be counted by microscopy with high accuracy. If high
cell densities are achieved, the dilution needed in order to count the
cells will however introduce too large an error to be of value. Several
types of counting chambers are available such as the Bürker chamber
(haemocytometer), where a selected area in the chamber corresponds
to a specific volume. Two factors are important: 1) the sample must be
diluted because too many cells cannot be properly distinguished and 2)
multiple samples from the same material must be counted to obtain a
reliable statistical estimation. Such microscopic images can be treated
by more or less complex techniques than the naked eye and they can in
more sophisticated equipment be captured on video and data-processed
by computer software. Figure 4.25: Coulter counter

To count a relatively small number of fairly large cells, the “Coulter


counter” type of equipment can also be used. This particle counter builds
on the high electrical resistance of cells. The cells are suspended in an
electrolyte and are passed, one-by-one, into a chamber through an orifice
over which an electrical current is applied. As they pass, they break the
current and this causes a drop in the voltage. Each pulse corresponds to
one cell. In addition, the size of the pulse is proportional to the size of
the cell, and this information is thus also obtained by this method. The
method is mostly used for counting mammalian cells such as white and
red blood cells.
Cells can also be counted on agar plates of various kinds. A sample
is diluted in a series of dilution steps and spread on the plates. After
incubation, the colonies can be counted (CFU: Colony Forming Units).
Only plates with a clearly distinguishable number of single colonies are
counted and this is hopefully introduced by the dilution series. This
analysis displays naturally only the cells that are able to grow in the
particular environment of the agar plate which may be difficult for cell
from a submersed cultivation. In practice, the method can thus only
provides an estimate of the number of “agar-culturable” cells present in
the sample.
Cells that are genetically modified by the use of antibiotic resistance
markers are often counted on agar plates with and without the specific
antibiotic. This also provides information about the number of cells that
still carry the designed genetic background.

Light scattering

The most common way to rapidly estimate bacterial growth is by mea-


suring the cellular ability to scatter light according to Lambert Beer´s
law. This is done by simple photometers using visible light. The “Optical
62 4 Growth of Cells

Density” or “OD” is generally measured at particular wavelengths and


to study the growth of E. coli wavelengths of approximately 600 nm is
commonly used. This method is however unreliable because any change
in morphology, such as the formation of cell aggregates, will also change
the light scattering property. Another influential factor is the presence
of air bubbles in the sample especially when the cultivation is heavily
foaming and the foaming is changing due to the addition of an antifoam-
ing agent. Under these conditions it is very difficult to obtain reliable
growth curves. Data obtained by different photometers can further not be
compared since the value is strongly dependent on the equipment. The
cells must in any case be diluted into the linear range of the photometer
(usually very narrow) before measurement. This dilution also contributes
to the error of the method. OD is thus only recommended when a rapid
real-time appreciation of the growth status is required.
One goal in growth analysis development has been to measure growth
directly in the fermenter to permit on-line data logging. Lights of different
wavelengths have been used, e.g. NIR (Near Infrared Spectroscopy),
where light is introduced and captured by optical fibres. Despite some
successful results, the technique has not yet been used very frequently in
the industry. A drawback has been the difficulty in measuring the very
high densities found in industrial cultivation.
Flow cytometry is an analytical method that is performed off-line and
makes it possible to accumulate a multitude of information about the
sample including an appreciation of variations within the cell population.
Figure 4.26: Principle of flow cytometry anal- This is possible since the sample is passed through a nozzle which
ysis
causes the cells to “line-up” before passing a light source (laser) whereby
they are investigated one-by-one (see also Coulter counter). The light is
scattered and captured by a detector at two different angles with respect
to the incoming light (forward scatter (2-15°) and side scatter (15-90°)).
The forward scatter carries information about the form, size, diameter
and volume of the cell and thus the cell number. The side scatter is often
used with cells that carry a fluorescent stain designed to mark a specific
feature of the cell.
This method enables a large number of cells to be counted in a relatively
short time as well as the detection of specific metabolic characteristics of
the individual cell. Examples of such information are: a colouring agent
binding to interesting molecules or giving rise to fluorescence as a result
of a particular reaction, changes in characteristics at a specific pH and the
localisation of antibodies which carry a fluorescent probe. The capacity
of uptake and also release of fluorescent agents over the cell membrane
can also be indicative of the membrane status and tell something about
cell viability.
Figure 4.27: Flow cytometry image of three Flow cytometry was developed for relatively large cells, such as cells
recombinant proteins expressed on the surface
of an E. coli cell. The proteins are coloured in of animal origin, but today parameters of small bacterial cells can now
green, blue and purpur, respectively. The red also be distinguished from the background noise thanks to equipment
image is a control i.e. the E. coli cell without
development.
protein expression

Weighing

Bacterial and fungal cell growth is often estimated by direct weighing


of a cell sample. For bacteria such as E. coli, often 3-5 ml, in triplicate
4.15 Measurements of medium and intracellular components 63

or quadruplicate, is placed in pre-weighed test tubes and the cells


are allowed to settle by centrifugation in a tabletop centrifuge. The
supernatant is discarded and the cells may thereafter be washed to
remove medium components and centrifuged once more. The cell pellet
is dried overnight at 105°C, allowed to equilibrate to ambient temperature
and weighed. This results in an estimation of the weight of the dried
cells, the “Cell Dry Weight” (CDW) with the unit [g/L]. Sources of error
are the discarding of cells with the supernatant and the interference of
salts or other nutrients present in the cell pellet.
A sample of fungal cells, on the other hand, is often filtered through
pre-weighed filter papers rather than being centrifuged. After washing
and drying, the papers are again weighed to give the cell mass.
Taken together, these different methods may give quite different estimates
of the growth rate since they are actually measuring quite different
parameters. Care must indeed be taken before choosing the analytical
method and deciding how to evaluate the data.
It is maybe not surprising that it is common in bacterial cultivation to
use a combination of methods. Optical density (OD) is an easy and fast
method to get a rough impression of whether and how the cells are
growing. This method thus allows an understanding of the process as it
proceeds. Simultaneously, measurements of the Cell Dry Weight (CDW)
are performed, but these results, although they are generally considered
to be the more reliable, are not available until the day after cultivation.
The results obtained by the two methods can then be compared. Below is
plotted the growth based on OD and CDW for comparison in a standard
E. coli cultivation.
In Figure 4.28, it is easy to see that the relation between OD and CDW
Figure 4.28: Comparison of growth curves
varies during the cultivation. The relationship is often a function of the based either on Optical density (OD) or on
growth rate so that only at a constant rate a stable quotient is obtained Cell Dry Weight (CDW) measurements in a
between the values. Which method should then be chosen to calculate cultivation of Escherichia coli

the specific growth rate? Generally, the error in the CDW is much smaller
and this method is therefore the more reliable.
The CDW is also needed to estimate the yield, a parameter that was
mentioned earlier, i.e. the ratio of the cell growth rate to the glucose
uptake rate. This “efficiency quotient”, i.e., Yx/s , is most easily expressed
in gram/gram, which means that the cell mass is needed in order for this
quotient to be calculated.

4.15 Measurements of medium and


intracellular components

The amounts of substrates, such as glucose or the nitrogen compounds,


are not often measured during industrial production. There are exceptions
and in waste water treatment it is of course important to keep track of the
degree of purification of carbon, nitrogen (Kjeldahl-N) and phosphorous.
However, during research and during process development a larger
number of measurements are often performed.
64 4 Growth of Cells

The actual technique used depends on the concentration. Low glucose concentrations can be measured by
prefabricated kits containing a set of enzymatic reactions which are commercially available. Another common
technique is the use of high pressure liquid chromatography, HPLC, which generally gives a better accuracy
and precision. It is here advisable to issue a warning. Before any type of analysis, care must be taken to get
a representative sample from the cultivation and the sampling technique must therefore always be carefully
scrutinised. Long sample handling times will radically alter the concentrations of both extra- and intracellular
components. The valves that are standard equipment for the sampling of bioreactors, generally take a very
long time to fully open (which is necessary to obtain a representative sample) and this is much too long to get
accurate results. For sampling of medium glucose, the time this can be allowed to take is easy to calculate
if the substrate consumption rate can be appreciated. For a rapidly growing E. coli culture, the sampling
time should generally be less than a second before reaching of the sample vessel. If reliable and absolute
values of the read-outs are needed, the cellular reactions need to be instantly stopped. Rapid freezing in
liquid nitrogen or use of dilute acid and ice-cold perchloric acid is often used for this purpose. If data on
intracellular metabolites are needed the sampling protocol will be even more critical since the turn-over is
very high. For measurements of such specific components, e.g. ATP and NADH, specific precautions need to
be added and the reader is referred to the literature.
Glucose electrodes are also available. Since these are mounted into the reactor they do not suffer from the
shortcomings of the sampling technique. Generally, these can however not be operated over the large glucose
concentration span in cultivation, with may range from values of 20 g/L down to a few mg/L, but they
can be used in important parts of these. Generally, however, the production industry is reluctant to insert
electrodes into the reactors due to the risk of breakage in the reactor and the contamination that follows and
further, the risk of compromising the sterility barrier.
Other cell-external analyses include measurements of overflow metabolites and products. The first category,
including compounds such as acetic acid, can also be analysed by the enzymatic reaction kits mentioned
above or e.g. by gas chromatography or HPLC. The product analysis is always developed on a case-by-case
basis and will therefore be determined by the actual product characteristics.
Yield and Productivity 5
When a new process is being designed, it is common to make a number 5.1 Yield . . . . . . . . . . . . . . . 67
of cultivations using different cell variants, different inoculum sizes, 5.2 Deviations from a constant
different medium compositions, different temperatures, different pHs yield . . . . . . . . . . . . . . . . . 69
Growth at very low glucose con-
and so on, to find the optimal conditions for the production of the
centrations . . . . . . . . . . . . . 69
product. At the same time, we also want to understand which parameters
The yield coefficient at high glu-
are particularly critical for the process outcome and the limits to which
cose concentration . . . . . . . . 73
these can be allowed to vary without any change in the outcome. We 5.3 Productivity . . . . . . . . . . 73
have also to keep in mind that there will mostly be two “products” to 5.4 The specific production rate 74
produce, i.e. both cells and the “true” product. To make a large total
amount of product we thus have to make many cells and each individual
cell, each “cell factory”, will also have to be extremely productive. When
we compare the data, we thus have to know something about the total
production (per reactor) or the volumetric production (per litre) as well
as the individual cell production i.e. the cell “specific production” (per
each cell factory). We note that all these three productivity terms are
rates (i.e. calculated per hour). The trial runs we perform may lead to
variations in the productivities listed above. It may then be that these
will diverge so that when a lot of cells are formed we get minor amounts
of product and vice versa.
In order to be able to compare the results from different conditions and
volumes and to understand which are to be preferred, we need to derive
some concepts that are valuable for this evaluation. It is here that the
concepts of “yield” and expressions regarding the concept “productivity”,
briefly introduced above, will appear. Before we go into further detail
of these concepts, we start by noting that certain particular letters are
commonly used to denote the interesting variables, listed in Table 5.1.
Please have a good look at the units!
Productivity is usually used for product formation while yield can be
used in many different variants. In the table, we see that three conversion
rate concepts are present i.e.:

I The biomass specific rate, where the letter "q" is used [mol/gx ,h]
Units of the rates can also be written as
I The volumetric rate, where the letter "r" is used. Sometimes written g/gx ,h, g/L,h and g,h, respectively
as the differential di/dt. [mol/L,h]
I The whole system rate, where the letter "R" is used [mol/h]

These three concepts are related, as shown by Equation 5.1

Definition 5.0.1 Relation of conversion rates a

𝑅 𝑖 = 𝑟𝑖 ∗ 𝑉 = 𝑞 𝑖 ∗ 𝑐 𝑥 ∗ 𝑉 (5.1)
a 𝑐 𝑥 = cell concentration
𝑉 = Volume
The suffix i indicates which substrate, product or cell mass is considered
66 5 Yield and Productivity

Table 5.1: Commonly used variables. Abbreviations: x: cells, s: substrate, p: product, o: oxygen.

Variable Abbreviation Unit


Cell mass Mx g or cmolx
Cell concentration cx g/L or kg/m3 or cmolx /m3
Substrate mass Ms g
Substrate concentration cs g/L or kg/m3 or mol/m3
Product mass Mp g
Product concentration cp g/L or kg/m3
Oxygen concentration co mmol/L or g/L
Byproduct mass (e.g. an overflow MHAc , MEtOH etc. g
metabolite)
Byproduct concentration cHAc , cEtOH g/L
Specific growth rate 𝜇 h-1
Specific substrate consumption qs mols /gx , h or gs /gx ,h
rate
Specific oxygen consumption rate qo molo /gx , h or go /gx ,h
Specific product formation rate qp molp /gx , h or gp /gx ,h
Volumetric biomass production rx cmolx /L, h or gx /L,h
rate
Volumetric substrate rs mols /L, h or gs /L,h
consumption rate
Volumetric oxygen consumption ro molo /L, h or go /L,h
rate
Volumetric product formation rp molp /L, h or gp /L,h
rate
Yield of cells per substrate Yx/s gx /gs or cmolx /mols
consumed
Yield of oxygen per substrate Yo/s go /gs or molo /mols
consumed
Yield of product per substrate Yp/s gp /gs or molp /mols
consumed
Yield of byproduct per substrate e.g. YHAc/s gbyproduct /gs or
consumed molbyproduct /mols
Total biomass production rate Rx cmolx /h or gx /h
Total substrate consumption rate Rs mols /h or gs /h
Total product production rate Rp molp /h or gp /h
5.1 Yield 67

5.1 Yield

In the more stringent mathematical form, the yield or “efficiency factor”, which we have introduced before, is
written as the quotient of two rates as follows:

Definition 5.1.1 Yield a


𝑅 𝑖 𝑟𝑖 𝑞 𝑖

𝑌𝑖𝑗 = = = (5.2)
𝑅𝑗 𝑟𝑗 𝑞𝑗

ai & j are any combination of cells, substrate, product, heat, metabolite etc.

The yield is thus always positive and the yield of any compound, whether substrate, product, overflow
metabolite etc, can always be related to any of the other parameters. The yield calculation is thus based on
the estimation of either the overall, volumetric or the specific rates, as shown in Equation 5.2.
To calculate the yield we can proceed as follows. If we wish to know the yield of cells produced in relation to
the consumption of glucose (and assuming constant volume), we start by stating the volumetric reaction
rates:

𝑑𝑐 𝑥
𝑟𝑥 = (5.3)
𝑑𝑡

𝑑𝑐 𝑠
𝑟𝑠 = (5.4)
𝑑𝑡

By combining according to Equation 5.2, we get:

𝑑𝑐 𝑥
𝑌𝑥/𝑠 = (5.5)
𝑑𝑐 𝑠

Integration from 𝑐 𝑥0 to 𝑐 𝑥 and 𝑐 𝑠0 to 𝑐 𝑠 , considering the yield as a constant, we finally get:

𝑐 𝑥 − 𝑐 𝑥0
𝑌𝑥/𝑠 = (5.6)
𝑐 𝑠 − 𝑐 𝑠0

from which the yield can easily be calculated if the cell mass and substrate concentrations are measured at
two separate time points. Since 𝑐 𝑠0 > 𝑐 𝑠 , the consumption will lead to a negative denominator. This does not
result in a negative yield because, by definition, the yield is always positive. The equation allows us both to
calculate the overall yield over the whole cultivation or the yield during any selected time interval provided
that it is constant.

Remark
Note Equation 5.6 uses biomass and substrate concentrations. For cultivations where volume is not constant,
volume change needs to be considered as well.

As a simple example, consider a medium with an initial concentration of 10 g/L glucose. After the glucose
is consumed and the cells are in the stationary phase, the measured cell dry weight is 5 g/L. Neglecting
the initial biomass concentration, this means that the yield is approximated at Yx/s =0.5 gx /gglucose . This is
more or less what we would expect from a theoretical calculation, provided that all the carbon goes either
into the cell mass or is used for respiration i.e. in line with the model of growth that we derived in the
preceding chapters. If we had found only 2.5 gx /L and a yield of only 0.25 gx /gs , we would then know that
something else is stealing the carbon. If this was an E. coli cultivation, we might guess that some carbon is
68 5 Yield and Productivity

being consumed for the formation of the common overflow metabolite for this cell, namely acetic acid. To
calculate the yields we may also use the volumetric or specific rates as seen in the yield definition above,
provided they are known. In our example of Yx/s we would thus use 𝜇 and 𝑞 𝑠 .

Table 5.2: Variance of Yx/s depending on nitrogen source and cultivation conditions.

Substrate/Nitrogen source Product Yx/s


Carbohydrate/Ammonia Cells 0.5
Carbohydrate/Amino acids Cells 0.7
Carbohydrate/Anaerobic Cells 0.1

The yield of cells from a given carbon substrate is often very interesting to know in process development
and depends on the organism and substrate in question. Table 5.3 shows yield coefficients for cell mass
accumulation based on either glucose or oxygen. The differences are due to the particular metabolism of these
organisms. The table shows the yield during aerobic conditions. Many cells can also grow anaerobically, but
under these conditions the yield of cells is much lower, since a much larger proportion of the carbon source
is utilised to generate energy. The yield with respect to carbohydrate consumption also depends greatly on
which nitrogen source is used and if oxygen is present or not (Table 5.2)

Table 5.3: Yield coefficients for biomass accumulation for a selection of microorganisms and substrates

Organism Substrate Yx/s [gx /gs ] Yx/o [gx /go ]


Enterobacter aerogenes Glucose 0.40 1.11
Fructose 0.42 1.46
Mannitol 0.52 1.18
Maltose 0.46 1.50
Glycerol 0.45 0.97
Succinate 0.25 0.62
Pyruvate 0.20 0.48
Saccharomyces cerevisiae Glucose 0.50 0.97
Methylococcus sp. Methane 1.01 0.29

The yield of carbon dioxide produced per oxygen consumed is called the respiratory quotient, or RQ. This
parameter shows whether there is a balance between these compounds i.e. if the consumption rate of oxygen
is in proportion to the production rate of carbon dioxide. This will not be the case if the cells are performing
reactions that lead to extensive carboxylation or decarboxylation and a calculation of the RQ can thus be used
to understand if either of these metabolic states are present.

Definition 5.1.2 Respiratory quotient


𝑞 𝐶𝑂2
𝑅𝑄 = (5.7)
𝑞 𝑂2

When is the yield important? Let us again take Yx/s as an example. If the maximum yield for the growth of
a particular cell on a substrate is known, we can easily understand whether the introduction of new operating
conditions alters the metabolism. It can be that the carbon flux is still solely directed to the formation of cell
mass or it may be that the process is partly facilitating the accumulation of other carbon-rich compounds.
Analyses of the actual sugar and the cell mass content will answer this. For some processes, e.g. cheap bulk
products like bulk chemicals and Baker´s yeast, the yield is especially important. In these processes, the
substrate cost is one of the largest parts of the production cost and care must be taken to ensure that all
nutrients go to product formation and do not (a) lead to anything else or (b) leave the process as waste. Here,
a yield, Yx/s , comparison is most important. Also in research and development phases, when cultivating on
5.2 Deviations from a constant yield 69

the laboratory scale, a Yx/s calculation can rapidly tell us whether we have reached a stationary phase due to
carbon substrate exhaustion or whether it is e.g. due to exhaustion of another nutrient.

5.2 Deviations from a constant yield

The calculation of the yield allows us to obtain an impression of how the flux of substrates takes place in the
cell and where the elementary components can eventually be found. From the definition above it is evident
that that the yield is actually expressing the relation between the fluxes of two parameters. The tables above
show that this relation greatly depends on the medium used. Apart from this, we can unfortunately not even
assume that the yield is constant over the time course of a single cultivation. During growth on e.g. glucose,
a very common carbon source, there are two cases where this assumption is probably not valid: at a very low
or at a very high concentration of glucose, as we shall see below.

Growth at very low glucose concentrations

In industrial processes, cells are often grown at very low glucose concentrations (the reason for this is
explained later in this book). This means that the glucose concentration is in the range of the Ks -value and
when there is so little carbon to use, this may constitute a problem for the cell to handle and we remember
that the metabolic model previously proposed was based on the assumption that the carbon taken up by the
cell is only used to create new cells and to generate energy.
However, it is maybe easy to understand that each cell also needs some energy for “housekeeping” activities
i.e. for survival. These can e.g. be related to sustaining the integrity towards the environment by maintenance
of the membrane potential. Based on such prior knowledge, we may therefore modify our previous model
with the claim that at low glucose concentrations, the maintenance requirements have priority over the
making of new cells. This means that at a sufficiently low glucose concentration (or glucose flux to the cell)
all the carbon is used for maintenance purposes and nothing may be left for the making of new cells. Let us
see how this affects the yield coefficient, Yx/s .
As stated above, the yield of cells per glucose is by definition:

𝜇
𝑌𝑥/𝑠 =
𝑞𝑠

If at low glucose consumption rate, growth would indeed not be possible, the yield (Yx/s ) can by definition
not be constant. This means that we need another expression describing the relationship between µ and qs .
To understand how this works, we design a cultivation experiment in such a way that we can make
simultaneous measurements of qs and µ as we reduce the glucose concentration. This cannot be made in what
we call “a simple batch cultivation”, since the drop in glucose concentration is so fast that we will not have
time to take samples. This rapid drop was visualised in the substrate consumption graphs and in the graph
describing the Monod expression in the former chapter on growth. For this purpose, we will therefore use
another cultivation concept which is described later on in this book and called “continuous cultivation”.
When we have finalised such an experiment, drawing of the straight line through our data points leads to
an intercept at a certain value on the y-axis (Figure 5.1). This value represents thus a certain rate of glucose
consumption which does not result in growth but is necessary for the “housekeeping” reactions. We will call
this substrate uptake: the “maintenance” value, which is denoted qm and has thus the same unit as qs i.e.
g/g,h.
If we rely on these data, we need thus to modify the expression for the substrate uptake (and yield) to
also be valid at low values. This is done by addition of a new term to our previous model representing the
maintenance glucose flux, resulting in the Herbert-Pirt equation (Equation 5.8).
70 5 Yield and Productivity

Figure 5.1: Definition of the concept of maintenance. The crosses refer to experimental data indicating by the red dotted line, that at a low glucose flux
into the cell, no new cells are made although glucose is being consumed. This is proposed to be used for maintenance purposes

Definition 5.2.1 The Herbert-Pirt equationa


𝜇
𝑞𝑠 = + 𝑞𝑚 (5.8)
𝑌𝑒𝑚
a in absence of non-dissimilatory products

Yem is the yield excluding the maintenance (represented by one divided by the slope of the curve). This
means that the yield goes down, especially at low 𝜇, since less glucose is actually used to produce the cells,
because some of it is being used to support the maintenance requirements.
This maintenance model, which was originally proposed by Prof. S. John Pirt, can be challenged by another
view. It can instead be suggested that the maintenance energy is internally generated by e.g. degradation of
the cell itself. This has often been detected in the production e.g. of recombinant protein, i.e. when protein
production was induced, the lack of substrate was shown to lead to degradation of the ribosomes to support
the immediate need for nutrients and energy. However, we will use the Pirt model suggesting an external
source in the discussion to follow.
The qm -value needs to be experimentally determined for each organism, but a rule-of-thumb is that this
value is 1-10% of the qs,max -value. Some maintenance values for different microorganisms under different
conditions are shown in Table 5.4. In this table we can also see the large effect on maintenance of addition of a
high concentration of salts, NaCl, in this case in yeast cultivation. The need for the cell to use energy to adjust
the membrane potential to balance this high concentration greatly increases the maintenance demand.

Table 5.4: qm -values for a range of microorganisms and conditions

Organism qm [gglucose /gcells ,h]


Escherichia coli 0.40
Saccharomyces cerevisiae (aerobic) 0.01
Saccharomyces cerevisiae (anaerobic) 0.036
Saccharomyces cerevisiae (anaerobic, 1M NaCl) 0.36
Penicillum chrysogenum 0.02
Aspergillus nidulans 0.62
Aerobacter aerogenes 0.48
5.2 Deviations from a constant yield 71

We may now refine our earlier cell model to include the maintenance requirements, where we note that the
flux to maintenance is used only to generate energy and not in the formation of new cells according to its
definition. In the model (Figure 5.2), qo , the specific oxygen consumption [g/g,h], is also included. This is to
underline the fact that the oxygen taken up by the cells from the air is used only for energy generation. The
oxygen going into the cell mass comes from the oxygen content of other nutrients.

Figure 5.2: Metabolic model of carbon flux inside the cell including maintenance requirements. Note that the oxygen consumption from the air is only
used for generation of energy

The models we now have adopted of the cell we use in order to increase the understanding of what is going
on on the inside and maybe later on to do something rational about cell metabolism in particular when
we want to introduce some kind of product formation in the cell. The present model we call a “structured”
model of the cell. This differs from a “black-box” model which only considers what is going in and what is
coming out without any interest in how this reflects the distribution amongst the reactions inside the cell.
One example of a black-box model is a stoichiometric balance of the substrates going in and products going
out of the cell, in which case we are only concerned with measurements of concentrations outside the cell.
The magnitudes of the fluxes in Figure 5.2 are not drawn to scale and are only used to show the principle. As
has been said before, approximately half the carbon flux goes to the creation of new cells and the other half to
energy generation, whereof 1-10% of the energy generated goes to maintenance.
If we want to calculate unknown fluxes inside the cell at this point, we may perform a mathematical simulation
based on these equations. We have to have some figures for the known fluxes (that we can easily measure)
but we may calculate the others. A balance is always made to include all the fluxes, in this case primarily of
carbon and for our present purpose the balance is made over the cell boundary. In later chapters we will set
up such flux balances also for other situations.
In order to set up the equations, let us for simplicity rename the carbon fluxes of the figure according to Table
5.5:
Table 5.5: Name conversion of variables used in Figure 5.2 for adaptation to equations

Term used in Figure 5.2 Term used in equations


qs, cellmass qsan where "an" indicates anabolism
qs, energy qsen where "an" indicates energy
qs, energy/growth qsen,g
qs, energy/maintenance qm

While we already have kinetic expessions for qs and qm we need to derive some expressions for qsen , qsan and
qo . The amount of carbon used to generate cell mass is, as previously stated, roughly equal to the total carbon
flux entering the cell minus that used for maintenance and this is denoted, qsan . This carbon flux further
equals the carbon actually found by growth of the biomass according to:

𝑞 𝑠𝑎𝑛 ∗ 𝑓𝑐,𝑠 = 𝜇 ∗ 𝑓𝑐,𝑥 (5.9)


72 5 Yield and Productivity

where the coefficients “fc,x [gCarbon /gcells ]” and “fc,s [gCarbon /gsubstrate ]” are the fractions of carbon in,
respectively, biomass and substrate, used to compare the carbon flux in between these two flows. The carbon
content in the substrate is easily determined but the content in the cells is not obvious, as we have stated
before. Although such figures can be found in the literature, these vary depending on the conditions under
which they were determined and only an elemental analysis of the cells under relevant conditions will allow
an accurate estimation.
We now substitute the growth rate using the Herbert-Pirt equation (Equation 5.8):

𝑔𝑐𝑎𝑟𝑏𝑜𝑛
 
𝑞 𝑠 𝑎𝑛 ∗ 𝑓𝑐,𝑠 = (𝑞 𝑠 − 𝑞 𝑚 ) ∗ 𝑌𝑒𝑚 ∗ 𝑓𝑐,𝑥 (5.10)
𝑔𝑐𝑒𝑙𝑙𝑠 , ℎ

This gives the final expression for the intracellular carbon flux to biomass, which can only be evaluated if the
maintenance requirement is known.

Definition 5.2.2 Intracellular carbon flux to biomass

𝑓𝑐,𝑥
𝑞 𝑠 𝑎𝑛 = (𝑞 𝑠 − 𝑞 𝑚 ) ∗ ∗ 𝑌𝑒𝑚 (5.11)
𝑓𝑐,𝑠

The intracellular carbon flux leading to energy generation, qsen , is then what is left from the total flux of
carbon and is given by subtraction of the amount going to biomass and the use of the expression for qsan
gives:

Definition 5.2.3 Intracellular carbon flux to energy generation

𝑓𝑐,𝑥
𝑞 𝑠𝑒𝑛 = 𝑞 𝑠 − 𝑞 𝑠𝑎𝑛 = 𝑞 𝑠 − (𝑞 𝑠 − 𝑞 𝑚 ) ∗ ∗ 𝑌𝑒𝑚 (5.12)
𝑓𝑐,𝑠

To derive the expression for qo we use the amount of oxygen required to fully respire the substrate (the yield
𝑟𝑒 𝑠𝑝
coefficient 𝑌𝑜/𝑠 ), which links the flux of carbon to energy with the oxygen flux according to:

𝑟𝑒𝑠𝑝 𝑞𝑜
𝑌𝑜/𝑠 = (5.13)
𝑞 𝑠𝑒𝑛

By combining Equation 5.13 and Equation 5.12 we get:

𝑓𝑐,𝑥
 
𝑟𝑒 𝑠𝑝
𝑞 𝑜 = 𝑌𝑜/𝑠 ∗ 𝑞 𝑠 − (𝑞 𝑠 − 𝑞 𝑚 ) ∗ ∗ 𝑌𝑒𝑚 (5.14)
𝑓𝑐,𝑠

When the carbon flux is very low a larger proportion of the energy goes to maintenance as we have stated
before. Since the supply of oxygen from the gas phase, qo , is utilised for maintenance and not for building
new cells, the relative amount of oxygen needed under such conditions will increase. In other words, the
relation between the oxygen and the carbon flux is not constant. At low substrate levels, the real yield
coefficient (𝑌𝑜/𝑠 = 𝑞 𝑜 /𝑞 𝑠 ) increases and more oxygen is needed to satisfy the proportionately higher qs flux to
maintenance.
5.3 Productivity 73

The yield coefficient at high glucose concentration

We have earlier talked about overflow metabolism as a situation where cells continue to take up a carbon
source even though they are not able to fully oxidize it to carbon dioxide and water. In such a case, E. coli
produces acetate, Saccharomyces cerevisiae produces ethanol and animal cells produce alanine and lactate.
When these products are produced in addition to the cell mass, it is thus clear that some of the carbon taken
up is incorporated into these products and less is used to create cell mass. This means that Yx/s , will be lower
when cells are producing these metabolites i.e. at high glucose concentration.

5.3 Productivity

We will now finally look at some valuable expressions by which we may try to understand the rate of product
formation.
How the product formation is described and which parameters that are used to do so, depends on the
process. In some processes, in particular when the product is very expensive, the concentration of the product
or the titer, is the preferred expression. For such products the quality is often more crucial than the quantity
and an example is the production of pharmaceutical protein. It is however the concept of productivity, on the
other hand, that for the first time introduces the rate of product formation, i.e. the parameter that shows how
much product we can actually produce per time unit. For most processes this is naturally a very important
parameter.
When is the productivity concept truly important? Well, in most processes, time is money, and a long-term
operation of the bioreactors is very expensive. It is thus an economical benefit if the process which leads to
the desired amount of product, is short. There is also a pressure from the customer and the market not to lose
shares. The total rate of production, i.e. the amount of product that can leave the factory every day (Rp ), or
the total productivity, is very important. If these amounts cannot be produced, then new and larger reactors
(and the corresponding plants) have to be installed which is a long-term development project.
To get a mathematical expression for the productivity we note that we have earlier described the specific
productivity concepts also for cell production and for the specific consumption of glucose. We recall that the
cell-specific productivity of new biomass was designated “𝜇” and the specific glucose consumption “qs ”. The
biomass-specific productivity of the product is consequently denoted, qp .
We note that these rates i.e. the volumetric, the specific and the total productivity can principally be defined
in a similar way for all these three parameters. We summarise these productivity concepts as follows:

𝑑𝑐 𝑥 1 𝑑𝑐 𝑥
𝑟𝑥 = 𝜇= ∗ 𝑅𝑥 = 𝜇 ∗ 𝑐𝑥 ∗ 𝑉
𝑑𝑡 𝑐 𝑥 𝑑𝑡
𝑑𝑐 𝑠 1 𝑑𝑐 𝑠
𝑟𝑠 = 𝑞𝑠 = ∗ 𝑅𝑠 = 𝑞𝑠 ∗ 𝑐𝑥 ∗ 𝑉
𝑑𝑡 𝑐 𝑥 𝑑𝑡
𝑑𝑐 𝑝 1 𝑑𝑐 𝑝
𝑟𝑝 = 𝑞𝑝 = ∗ 𝑅𝑝 = 𝑞𝑝 ∗ 𝑐𝑥 ∗ 𝑉
𝑑𝑡 𝑐 𝑥 𝑑𝑡

During process development we want to optimise both the cell and the product concentration. If the
cultivations have been run to the same cell mass it is principally sufficient just to compare the product
concentration. This situation is however very rare and the cell mass will differ between the various cultivation
setups. The product formation is thus best approximated by the specific productivity i.e. the product per unit
cell mass. This parameter can be thought of as an optimisation of the “cell factory” productivity i.e. it is a
measure of what a single cell can actually make. Many factors of both genetic origin and associated with
the selected micro-environment will affect this parameter and this is studied in the initial phases of process
development.
74 5 Yield and Productivity

If different reactors or reactor volumes have been used, on the other hand, the volumetric productivities, i.e. the
production per litre culture volume, must be compared as well as the specific productivities. And finally,
when the study has been scaled-up to production size, we need to know the total productivity i.e. how much
the factory can totally produce in unit time (Rp ) and if this is enough to cover the market share.
To understand the concept of specific product formation better we will look at an example. The first graph
in Figure 5.3 shows the product accumulation with time from two different experiments with the same
cell (“product 1 and 2”). These can e.g. differ due to different media or different modifications of the cell.
In this graph it is evident that the product accumulation [g/L] is more rapid in experiment (1) and the
conditions of this cultivation may therefore be the natural first choice. We note also that the accumulation
rate is exponential which means that the product is produced along with the cell growth.
In order to decide which of the two processes that should be selected for continued development and to get a
clue to which way this process could be further improved, it may be important to consider other data than just
the product concentration curves in particular since we here have a growth coupled production. A good idea
is thus to check the cell growth. The cell growth data are displayed in Figure 5.3 where it is obvious that
growth is considerably better under the conditions of experiment (1).
600 250 0.3
SPECIFIC PRODUCTION RATE (gP/gx,h)
CELL MASS ACCUMULATION
500 PRODUCT FORMATION 0.25
200

400 0.2

qP (gP/gx,h)
Cell concentration 1 (g/L)
cP (g/L)

150
cx (g/L)

Cell concentration 2 (g/L) Product 1


Product 2
300 0.15

100
200 0.1

50
100 0.05
Specific rate 1
Specific rate 2
0 0 0
0 2 4 6 8 10 12 14 16 0 2 4 6 8 10 12 14 16 0 2 4 6 8 10 12 14 16
Time (h) Time (h) Time (h)

Figure 5.3: Time-dependent accumulation of product (left) and cell mass (middle), and development of the specific production rate (right) in
cultivations 1 and 2

Since the experiments however result in different cell concentrations it might be a good idea to plot the
specific rates to get some information about the productivity of each cell and here it is obvious that the
specific productivity is much better in experiment (2), where each cell produces much more product despite
the cell mass being lower.
Further optimization studies might thus focus on increasing the cell mass under the conditions of experiment
2 rather than on changing the parameters that control the production which might be considerably more
difficult if they have to take place on the cell design level. We can also conclude that in experiment (1) the
specific productivity decreases rapidly at the end of the cultivation. This is a clear disadvantage since this
is the point at which we shall harvest the cells and isolate the product and we do not want to do so under
unstable conditions since the cell harvest takes some time and this declination can lead to an unstable and
unpredictable outcome.

5.4 The specific production rate

This is in any case the starting point. In the previous example we saw that the product formation was coupled
to the cell growth since both parameters were exponentially increasing. But there are in reality many ways by
which a product may be formed. However, some basic models are mainly used which reflect some different
and well-known product categories. We start from stating that product formation rate models are almost
always looked upon from their dependence on the specific growth rate.
The first group (1) consists of the particular cell metabolites that are accumulated simultaneously with the
cell growth and which are thus called “growth associated”. The relation between growth rate and product
5.4 The specific production rate 75

qP [gP/gx,h]
qP [gP/gx,h]

Mx [gx]

MP [gP]
MP [gP]

μ [h-1]
Mx [gx]
μ [h-1]

Figure 5.4: Selected types of production kinetics during batch cultivation. Left: growth coupled production. Right: Mixed growth coupling
(Luedeking-Piret kinetics) Legend. Black: specific growth rate [h-1], blue: cell accumulation [g/L], green: product accumulation [g/L] and red: specific
production rate (g/g,h)

formation can in this case be given by the yield of product per biomass produced. This yield is then taken as
constant over the cultivation. Such a production pattern might apply to the production of primary metabolites
where amino acids can serve as a good example. The second (2) principal product accumulation pattern
appears when there is no clear-cut dependence of the product formation on growth. Product is formed both
in relation to growth but also in some amounts as uncoupled to this. The “Luedeking-Piret” model has
often been used to describe this type of kinetic performance. The accumulation of product based on the two
different models over time is shown in Figure 5.4.

Definition 5.4.1 Different forms of product formation rate

1 𝑑𝑐 𝑝
𝑞𝑝 = ∗ = 𝑌𝑝/𝑥 ∗ 𝜇 𝑔𝑟𝑜𝑤𝑡 ℎ 𝑐𝑜𝑢𝑝𝑙𝑒 𝑑 (5.15)
𝑐 𝑥 𝑑𝑡

𝑞𝑝 = 𝛼 ∗ 𝜇 + 𝛽 𝑚𝑖𝑥𝑒 𝑑 𝑔𝑟𝑜𝑤𝑡 ℎ 𝑐𝑜𝑢𝑝𝑙𝑒 𝑑 (5.16)

𝑞 𝑝 = 𝑐𝑜𝑛𝑠𝑡 𝑎𝑛𝑡 = 𝛽 𝑔𝑟𝑜𝑤𝑡 ℎ 𝑢𝑛𝑐𝑜𝑢𝑝𝑙𝑒 𝑑 (5.17)

In mathematical terms the production rates can for example be expressed as Equation 5.15 and Equation
5.16 where Yp/x , 𝛼 (a growth coupled yield) and 𝛽 (a production rate when there is no growth) are used to
approximate the production kinetics. Equation 5.15 is used to describe the growth-associated production
and Equation 5.16 the combination of a growth coupled and a constant non-growth-associated production.
Equation 5.16 is the “Luedeking-Piret” equation where 𝛼 is the yield during growth.
The third characteristic type of product formation (3) occurs when there is no connection to the growth at
all. This means that production occurs only after growth has ceased. The specific production, qp , is thus
zero during growth. If this production is constant after growth has ceased, we could express the specific
productivity as Equation 5.17. This production pattern is typical e.g. for the production of penicillin and also
other for antibiotics and other molecules which are called “secondary metabolites” to distinguish them from
the “primary metabolites” formed in connection with growth.
If these metabolites are not formed when the cell grows one can perhaps ask the question: how can we then
produce them? The answer is that we produce under very low supply of nutrients and not under starvation.
For this, we have to apply other cultivation concepts than batch processes: the so called “fedbatch cultivation”
or “continuous cultivation” concepts are used. More theory of these concepts are given in a later chapter but
the “trick” used to get this type of production going is to use a very slow feed of sugar (and/or nitrogen)
after the batch phase is finished i.e. when the initial sugar has been consumed. This reduces the maximum
growth rate considerably and production can start.
76 5 Yield and Productivity

qP [gP/gx,h]
Mx [gx]

MP [gP]
μ [h-1]

Figure 5.5: “Fedbatch” cultivation for the production of a “secondary metabolite” i.e. a metabolite produced after growth has ceased. Black: the specific
growth rate (h-1 ), blue: cell mass accumulation (g/L), red: the specific production rate (g/g,h) and green: product concentration (g/L). Only when the
specific growth rate is close to zero, the production can start!

The principal outline of a fedbatch cultivation for this type of product can be found in Figure 5.5, where the
rate of production is approximated as constant, once it starts. Here we can readily see that the product is only
produced when the nutrient concentration is limiting and the specific growth rate very low. This reduces the
cell mass formation, which is here only linearly increasing in favour of the product formation.
Bioreactors 6
A bioreactor can be defined simply as a vessel in which a bioreaction 6.1 The stirred tank reactor . . . 78
is performed. Many such vessels have been developed but only a few 6.2 Gas-mixed bioreactors . . . . 80
have come to stay. In the beginning of this book it was mentioned that a 6.3 Other types of bioreactors . 82
Packed beds . . . . . . . . . . 82
bioreactor can also be called a “fermenter” and this term is often used
Wave bioreactors . . . . . . . 83
although the cultivation is aerobic. On a small scale, agar plates, micro-
6.4 Measurement and control of
titer plates and shake flasks could, by this definition, also be considered
bioreactors . . . . . . . . . . . . . 83
as bioreactors. However, this is rarely done and what we consider to be a In situ analysis . . . . . . . . . 84
bioreactor is rather a dedicated vessel which is designed, sold and used Liquid sampling for external
only for the purpose of cultivation and which will live up to a set of analysis . . . . . . . . . . . . . . . 85
qualifying demands regarding its performance. Gas sampling . . . . . . . . . 85
Supervising the bioreactor per-
I Functionality - It should do its job! formance . . . . . . . . . . . . . . 85
I Reasonable investment cost - Set in relation to the product price
I Aseptic performance - Competing cells should not grow
I Easy cleaning & access - e.g. ports for addition and removal of
compounds
I Automation - Allow for measurement & control of important
parameters

A few dedicated companies commercially design and produce fermenters


worldwide and they also provide service contracts for maintenance of
the equipment. Apart from the actual bioreactor, computer-controlled
software and an interface between these are also part of the bioreactor
set-up as is equipment for analysis and control.
The smallest bioreactors, with volumes from a few hundred millilitres
up to a maximum of about 10 litres are usually made of glass. Such
bioreactors are used in process development and sometimes also as a step
Figure 6.1: Standard laboratory-scale bioreac-
in the scale-up to the production scale. Such bioreactors are sterilised in tors. These are all stirred tank reactors (STR).
an autoclave for quite a long time before they can be used in operation. See text for further explanation
In spite of their small size they can be fully automized.
During process development, a set of parallel and equally equipped
reactors of the same size is often used. This saves time because they can
be run simultaneously but with different parameter settings. These are
often fully automized.
The industrial scale bioreactor is always a stainless steel vessel. The
steel is of high quality and the surfaces very finely polished to avoid
corrosion and to avoid microbial growth on the surface. The volume
of a commercial scale bioreactor depends on the annual volume to be
produced by the company. For very specific biopharmaceuticals the
bioreactor may be only a few hundred litres in size, whereas the volume
of the Swedish bioreactor for Baker’s yeast production is over 200 m3 .
The latter bioreactor, being 14 m tall, occupies a factory from floor to
roof. Figure 6.2: A set of 2*six fully automated
parallel stainless steel 1 L bioreactors for pro-
The large-scale commercial vessels are sterilised if this can be afforded; cess development and sterilised in-situ, Belach
Biotechnology AB, Sweden
in other cases they are only thoroughly cleaned. In the case of large
78 6 Bioreactors

bioreactors, the CIP (Cleaning-In-Place)-procedure involves washing the


inside of the bioreactor with e.g. hot diluted acid or alkali dispersed
through specially designed equipment followed by rinsing with water. If
possible, the large-scale fermenters are of course preferable sterilised or
cleaned in situ. The sterilisable bioreactor is basically a vessel that can be
pressurised i.e. it has a double wall. Hot water or steam can be introduced
into the intermediate space, the jacket, to increase the temperature during
sterilisation. To increase the pressure, air is introduced while all ports to
the reactor are closed.
Bioreactors can be sterilised with the basic medium components in place,
but this is only done at bench- or laboratory-scale. The large-scale reactor
is first cleaned and sterilised as an empty vessel after which the pre-
sterilised medium components (see the chapter on media) are pumped
into the reactor. The temperature is then set to the desired operating
value and controlled by the addition of steam or hot/cold water. These
media are added to the jacket or to internal sets of stainless steel coils
in the bioreactor or both. All parameters that are deemed necessary for
control of the bioreactor are calibrated before sterilisation but rechecked
once the temperature has reached the set point. Thereafter, the operating
value set-points are set. Computers are automatically controlling this
procedure by from beginning to end. When all the parameters have
reached their steady state the bioreactor is inoculated.
There are three major types of bioreactor configurations that have come
to stay and these are all selected to suit particular processes. The three
types are:

I Mechanically Stirred Tank Reactor (STR)


I Bubble column reactor
I Air-lift reactor (special case of bubble column)

These reactors all operate with the cells in a liquid suspension which is the
most common type of bioprocess and where the liquid is mechanically
stirred or stirred by introduction of compressed air or both. These
bioreactor characteristics are described in more detail below.

6.1 The stirred tank reactor

The stirred tank reactor (STR) consists of a stainless steel vessel which
is stirred by mechanical stirrers called agitators or impellers that are
installed in the reactor. In Figure 6.3, showing the STR top, the stirrer
is mounted from the top where the motor, gearbox and the sealing are
placed. It is however equally common that the impeller is fitted at the
bottom of the reactor often by a magnetic sealing. In large vessels, several
such stirrers are generally mounted on a shaft on top of each other. These
Figure 6.3: The top of a pilot-scale bioreactor stirrers can be of different configuration and this affects the efficiency of
that is stirred from the top. The bearing can
be internal or external; the connectors connect mixing (see chapter on mixing).
different items between the inside and outside
and are sealed to provide sterility. A sight glass When a reactor is stirred from the centre like this, a vortex is created
can be found on the side or at the top, also in in the middle of the reactor. To reduce this, a set of thin and narrow
very large bioreactors, and this makes it pos- vertical stainless steel-plates, “baffles”, are placed from top to bottom
sible to actually visualise the cell suspension.
Bearings and connectors are often steam-sealed close to the sides in the reactor. The number depends on the vessel but is
to protect against infection usually between 4 and 8. To supply oxygen to aerobic processes, clean or
6.1 The stirred tank reactor 79

sterilised air or in some cases pure oxygen is supplied by a compressor


and is introduced by a “sparger” which is mounted on a a pipe ending
with some kind of technical solution for distribution of the air. This is
often a ring-shaped, perforated stainless steel device placed below the
bottom stirrer since the bubbles, which have a lower density than the
liquid, will then consequently rise through the reactor and are further
distributed by the impeller. The outlet gas passes through a filter and
finally a condenser to avoid stripping the reactor of water.
Careful attention is taken with respect to the reactor proportions in order
to:

I Enusre that the contents are carefully mixed to avoid concentration


or temperature gradients
I Efficiently transfer gases over the gas-liquid interface to provide
the desired level of oxygenation and to remove carbon dioxide
(aerobic processes)

It is commonly understood that the reactor dimensions will have an


impact on, and also describe, the functionality of a bioreactor. The
abbreviations used in Figure 6.4 can be seen in Table 6.1. Not only
Figure 6.4: Principle outline of a stirred tank
the absolute dimensions but also the relationships between them are
reactor. See text for definitions of the abbrevia-
important for best operation. Some of these can also be used when a tions. The blue plates at the sides of the reactor
bioprocess is scaled to a larger size to keep geometric similarity and thus indicate the baffles. See the table for further
interpretation
functionality. In proportion, as compared to the other types of reactors i.e.
the “air lift” reactors, the STR is a rather short and “fat” vessel. Common
relationships for STRs are listed in Table 6.2.
Table 6.1: Commonly used abbreviations for
The stirred tank reactor has the highest functionality, as defined above, of bioreactor dimensions
all reactors but it is also the most expensive type since the costs for good
Abbreviation Description
blending and efficient oxygen transfer, achieved by the mechanical stirrer
HT Bioreactor height
and the high energy input, are very high. This reactor can thus never be
Di Impeller diameter
used for a product that cannot carry this price. The fact that it carries
DT Tank diameter
many internal components of the stirring device may be a potential threat
to sterility due to the many surfaces and the connections to the outside. Db Baffle width

A STR is however very commonly used since it is the most flexible Distance between
Ds tank bottom and
construction for a large variety of processes and process configurations. If lowest impeller
a company expects to run a number of different processes or expects the Distance between
𝛿D
present process to change in the future, a STR may be a good choice. impellers

Table 6.2: Common ratios for STR dimensions

Ratio Common value Comment


HT /DT 3-5 Max 2 for animal cell fermenters
DT /Di 3-4 For Rushton type impellers
DT /Di 1.7-2 For Hydrofoil type impellers
DT /Ds 3
Di / 𝛿 D 0.7-1 Depends on viscosity of cell suspension
DT /Db 10-12.5

Bioreactor operation depends not only on the reactor configuration and


geometry but also on other reactor parameters. These physical factors
are listed in Table 6.3 where some are indirectly coupled to the biology
of the process.
80 6 Bioreactors

Table 6.3: Common parameters influencing the efficiency of reactor operation

Factor Abbreviation Unit Comment


Power input P W, J/s, Nm/s Energy from motor
3
Power input per volume P/V kW/m
Energy from impeller dissipated
Energy dissipation 𝜀𝑇
to the liquid
Tip speed 𝑣 m/s Speed of impeller tip
Standardised expression for im-
RPM N Rotations per minute
peller rotation speed
Medium density 𝜌 kg/m3
In some literature other abbrevia-
Medium viscosity 𝜂 Ns/m2
tions are used
Gas velocity Q
Standardized expression for rate
VVM Q/V m3 gas/m3 liquid, min
of gassing
Gas flow over cross-sectional area
Superficial gas velocity vg m/s
of reactor
Ability of the medium to hold the
Gas holdup 𝜀 m3 gas/m3 suspension
gas

Many of these values will depend on the scale of cultivation and typical
values are (STR):

Table 6.4: Typical values for key parameters in different scales

Parameter/scale Laboratory scale Pilot scale Large scale


RPM 0-1500 0-500 0-50
Power input (STR)
8-10 3-5 1-2
(kW)
VVM (STR) 1-2 1 1
Oxygen transfer
400 intermediate 100
rate (mmol/L,h)

Table 6.4 clearly shows that values cannot be maintained when we are
going from the small to the large scale. This is therefore an important
factor to recognise before scaling-up a bioprocess to the production scale.
A too high aeration rate is often avoided although the equipment might
provide this. The reason is that a high rate promotes foaming, which is
very undesirable.

6.2 Gas-mixed bioreactors

In a gas-mixed bioreactor the main bioreactor functions, mixing and


aeration, are supported by the addition of large quantities of air. Two
main design principles are being used: the airlift and the bubble column.
The key to understanding the driving forces for the gas and liquid
flow in these reactors is that the sparging of gas at the bottom leads
to a density inversion in the gas-liquid dispersion which immediately
induces convective flow. Bubble columns are thus almost exclusively
6.2 Gas-mixed bioreactors 81

used in aerobic processes. The benefit with this reactor type is that the
design permits construction of very large scale reactors, which gives a
high power input per volume.
A bubble column reactor has a very basic and simple design and this is
very much on purpose. The bubble column will be used for large volume
products which are depending on good aeration but where infection must
be avoided and the internal equipment should therefore be minimised.
The principle outline is shown in Figure 6.5.
The bubble column is essentially a highly polished, stainless steel cylinder
where large amounts of compressed air are introduced at the bottom and
leave the reactor at the top. The air sparger can be of various in-house
designs and is generally a set of perforated pipes or plates, where the
larger the holes the larger are the bubbles. The liquid generally ascends
in the middle part of the reactor and flows down at the sides. The airflow
drives both the mixing and the oxygenation and a long residence time of
the air in the liquid is sought. The bubble column has typically a high
height-to-volume ratio of HT /DT = 4-6, giving a much taller and slimmer
reactor than the mechanically stirred tank. Not only the design of the
reactor and sparger but also the rheology of the broth i.e. the medium
viscosity determine the efficiency of the mixing and oxygen transfer.
The airlift bioreactor is slightly more complex than the bubble column
Figure 6.5: The basic design of a bubble col-
since a stationary mechanical device is installed to enforce the circulation umn reactor
and thus to improve the mixing. However, any such internally mounted
equipment can be a site for growth of microorganisms, which may not
be efficiently removed during cleaning and this needs to be considered
in relation to the sensitivity of the actual process.

Figure 6.6: Common types of gas-mixed bioreactors: left: internal draft tube, middle: baffle
split cylinder and right: external loop. The picture is drawn to show the principles and not
to exact proportions. At the top of the figure a cross section of the middle part of each
reactor is shown

The design divides the reactor into two main parts: a “riser” part for
the up-going air/medium and a “downcomer” part for the downward
flowing liquid that is stripped of the gas since this exits at the top. This
means that the density is much lower in the mixed gas/liquid part of the
riser and this medium will thus be driven upwards (the air lift). When
the gas is stripped from the medium at the top, the broth will become
denser and will consequently sink down towards the bottom. The rate
of circulation is roughly equal to the square root of the riser height.
This forced circulation is used when the blending of the broth is very
82 6 Bioreactors

important. The proportions of the two parts are approximately: DR /DT


= 1.8-4.3. However, since a considerable part of the reactor is ungassed,
the total oxygenation is not as good as in the bubble column.
Since the airlift and bubble column type of reactors contain no moving
parts, the energy cost is reflected only by the cost of the compressed air.
The comparatively lower price and the possibility of effective cleaning
the reactor without sterilisation makes these types of bioreactors advan-
tageous for the production of cheap products such as baker’s yeast and
citric acid.
The power input to STRs will be discussed later on in the compendium
but the energy needed for airlift bioreactor [W] can be approximated by
the gas flow and the pressure drop:

𝑃 = 𝑄 ∗ Δ𝑝 = 𝑣 𝑔 ∗ 𝑚𝐿 ∗ 𝑔 (6.1)

where Δ𝑝 is the pressure loss of the gas across the column, 𝑚𝐿 is the
liquid mass and g is the gravitational acceleration. To compare different
reactors, the power per volume, P/V, is a better parameter to use and
this results in the following expression:

𝑃
= 𝑣𝑔 ∗ 𝜌 ∗ 𝑔 (6.2)
𝑉

where 𝜌 is the density of the dispersion. Airlift bioreactors are mostly


operated at P/V = 1-2 [kW/m3 ]. The density is of course largely dependent
on the volume of gas that is dispersed in the liquid (the gas hold-up).

6.3 Other types of bioreactors

For some particular processes, other types of reactors are designed. This
is sometimes done to overcome a certain bottleneck in the production
but they are also often used as a result of tradition.

Packed beds

A solid bed reactor is a bioreactor in which the catalyst, cell or enzyme, is


immobilised on a solid support. This is used for example in the activated
sludge process in smaller wastewater treatment plants and traditionally
also for the production of acetic acid. The solid bed is also used for
many enzyme transformations and about half the worldwide biocatalysis
operations use a bound catalyst. The bed is thus made of particles on
which a biofilm of microorganisms grow or to which enzymes are coupled.
The medium is sprinkled over this bed, generally from the top, and air is
introduced in a counter-current flow with respect to the medium.
The bed material on which the microorganism grows is different for
different processes. In the “trickling filter” bioreactor for wastewater
treatment crushed stone is often used. In the early acetic acid process,
Figure 6.7: Packed bed reactor
birch shavings were used, and in the production of pharmaceutical
products the beads are made of a suitable and carefully designed and
6.4 Measurement and control of bioreactors 83

expensive polymer material. In many cases the processes run in this way
are highly oxygen-dependent and rely on the creation of a large surface
area which is provided by this reactor type.
Some cells cannot grow in suspension and the typical example is some
types of animal cells. Such cells are called “anchorage-dependent” and
they are often grown in membrane bioreactors particularly in “hollow
fibre reactors”. Here, a fibre surface is densely packed and a liquid
suspension is passed in channels between the surfaces. This allows Figure 6.8: Membrane bioreactors
the cells to grow on the fibers to a high density while the medium
continuously passes by. Such reactors can be used for many types of cells
and they have also been used sometimes in the purification of wastewater.
Problems are often encountered due to an uneven growth of cells over
the surface and partial clogging may thus occur when the supply of
oxygen rapidly becomes a limiting factor.
Another extreme example of a packed bed reactor is the “solid state”
bioreactor where the bed material is only kept “moist” and where little
or no liquid phase is present at all. Starter cultures for e.g. soy and
sake production (koji) are produced in solid state bioreactors where the
solid starting material is steamed rice which is inoculated with spores
Figure 6.9: Schematic drawing of a Wave
of Aspergillus oryzae. The growth of this mould leads to the excretion bioreactor showing the rocking motion which
of amylases and proteases that degrade the rice starch and make the prevents the settling of the cells.
monosaccharides accessible for processing by other organisms.

Wave bioreactors

Wave bioreactors are principally only used for animal cell cultivation.
These are in principle an advanced, inflated plastic bag where the medium
is slowly rocked back and forth to create a wave motion inside the bag
and in this way create the mixing.
The reactor is disposable, i.e. thrown away after use to avoid the always-
present risk of contamination in these processes. Animal cell cultivations
are particularly sensitive to contamination since the media are very rich
and these cells can easily be out-grown by many commonly present
microorganisms which grow much faster. Neither the gas-liquid transfer
nor the mixing is particularly good in these reactors but since the oxygen Figure 6.10: Large scale wave bioreactors. Top:
1-25L reactors. Bottom: 100-500L reactors. The
demand of animal cells is not as high as that of microorganisms, wave latter have the dimension: 2*2*1,2m
bioreactors can still permit a good reactor performance.
In the text above, the principles of the different bioreactor have been
shown. Apart from the reactor itself, a large amount of auxiliary equip-
ment is needed to run the cultivation and to supply the different media
that are continuously added. The framework around a bioreactor can
therefore look quite busy. This is indicated for a large-scale process in
Figure 6.11.

6.4 Measurement and control of bioreactors

Some selected parameters are always measured in bioreactors and the


desired value is usually also controlled. Although this differs depending Figure 6.11: Bioreactor and auxillary eqip-
on the product/process we sometimes talk of a minimum reactor ment. The Coscata process.
84 6 Bioreactors

Figure 6.12: Mimimum set-up for the measurement and control of a bioreactor. M: measured, C: controlled parameter. The dotted lines show how the
DOT signal can be used to steer either the ingoing gas flow or the RPM

standard that in most cases is needed to perform a cultivation process. The minimum standard can be
distinguished because many industries stop their investments at this level. The parameters that are usually
measured are the following:

I Temperature (sensor)
I Reactor pressure (sensor)
I pH (electrode)
I Liquid inflow and/or outflow (mass flow meter)
I Gas flow (mass flow meter)
I DOT (Dissolved Oxygen Sensor) i.e. liquid-phase oxygen content (electrode)
I Liquid level (conductivity sensor)
I RPM (impeller rotational speed)

These are all generally feedback-controlled to a “set-point” initially set by the process engineer as decided
during process development. The computer will then regulate this variable according to the accuracy and
precision allowed by the equipment. An exception is the oxygen level which is usually only monitored.
The pressure is particularly important if the reactor is sterilised and this is controlled by restriction of the
outgoing gas flow. A small over-pressure (0.1-0.2 bar) is often set in the reactor during cultivation to ensure
that nothing is drawn into the reactor from the outside which is ensured by the negative pressure difference.
The control algorithms are set and the signal regulated by computer-based automation. The resulting data
are logged at frequent time intervals.

In situ analysis

On-line (in situ) sensors are used for the measurement of: pH, temperature, liquid level and dissolved oxygen
(DOT). Before the cultivation is started, these are calibrated. Since these sensors penetrate the sterility barrier
over the bioreactor, they need to have distinct characteristics before they are introduced. They need to be:

I Sterilizable
I Stable over a long period of time
I Applicable over a large concentration span
I Inserted without breaking the sterile barrier
I Insensitive to protein adsorption and cell growth onthe surface
I Resistant to degradation (especially enzyme-based sensors)
I Without risk for the introduction of harmful and/or contaminating substances
6.4 Measurement and control of bioreactors 85

The sensors used today in the industry all fulfil most of these requirements. The adsorption on the surface is
however always a difficult task.

Liquid sampling for external analysis

Liquid samples are drawn from the bioreactor at intervals for off-line analysis of parameters that cannot be
analysed in situ. This is particularly the case for the analysis of cell mass and product which are almost never
analysed on-line. A specific sampling port is therefore introduced close to the bottom of the reactor. This is
often a three-way valve where one part is connected to the inside of the bioreactor, one to a flow of steam and
one to the outside. When a sample is taken, the connection towards the steam and the operator is first opened
to sterilise the passage. Then the connection between operator and bioreactor is opened and some material is
withdrawn and discarded due to the “dead volume” in the valve and then the actual sample is taken. After
sampling, the connection is again sterilised with steam to avoid contamination.

Gas sampling

In many bioreactors the outgoing gas is connected to an analyser. In this way, a continuous analysis of the gas
can be achieved. This can give information about many important parameters and is most often used for
measurements of the rate of respiration which is based on oxygen and carbon dioxide content of the outgoing
gas.

Supervising the bioreactor performance

As we have noted before, most bioreactors comes with a software which makes it possible to carefully
control the process to desired values. This is a true benefit compared to cultivation in, for example, the
unsupervised shake flasks where the environment of the cells is constantly changing. These changes can
seldom be monitored since the small volume of such flasks does not permit the withdrawal of many samples.
This means that this type of process is rarely reproducible and can only be used to give an indication of the
process performance.

Figure 6.13: Computer window for continuous monitoring of the bioprocess. Real time data are displayed to the very left in the figure. Set-points are
found to the immediate right of these
86 6 Bioreactors

Apart from control and data-logging, the software will also permit the plotting of all measured variables and
the time course of important parameters can be followed, and changed!, in real time. An example of such a
display is seen in Figure 6.13 where the real time values of reactor “FE1” are shown to the far left, the set
points to the right of these and the parameter label further to the right. The plot window is also displayed
along with the legend which is in the front of the figure.
Agitation 7
In a bioreactor, three physical phases are present: the liquid phase i.e. 7.1 The use of balance equations 87
the medium, the gas phase consisting of air which in aerobic processes 7.2 Macroscopic momentum balance
supplies oxygen and removes carbon dioxide, and the solid phase consti- describing the macromixing . . 89
7.3 Molecular transport balance de-
tuting the reactor hardware (and principally also the microorganisms).
scribing the macromixing . . . 90
Both the gas and the liquid phases are transported at a certain rate in the
7.4 Mixing devices . . . . . . . . 94
reactor mainly due to the use of particular mixing devices. The agitation
Agitator types . . . . . . . . . 94
of the bioreactor has thus two goals: firstly, to mix the liquid phase to 7.5 Turbulence . . . . . . . . . . . 96
ensure a homogeneous dispersion with no concentration gradients and 7.6 Measurements of mixing . . 98
no “dead spaces” where no mixing takes place; and secondly, to get a
good dispersion of the gas in the liquid to have a high mass transfer be-
tween these two phases. This will ensure the best possible aeration of the
broth where the microorganisms are situated. In this chapter, agitation is
mostly discussed from the point of mechanically stirred reactors. In the
final chapter on aeration, bubble columns are discussed.
Mixing takes place by transfer of momentum (Swedish: rörelsemängd),
which is one of three major transport processes:

I Transfer of momentum
I Transfer of mass
I Transfer of heat

The subject of this chapter is the transfer of momentum. The transfer of


mass will be discussed in the following chapters but the energy transfer
in bioreactors is not further treated.
When we wish to understand the efficiency of the transfer processes
we will the use so-called “balance equations” for the calculations. It is
possible to use balance equations for both momentum, mass and heat.

7.1 The use of balance equations

In classical mechanics we learn that momentum is the product of mass


times velocity, (m*v), with the units [kgm/s or Ns]. Momentum is a vector
quantity, which means that it has a direction as well as a magnitude.
Newton´s second law states that a body subjected to a force “F” undergoes
acceleration in the same direction as the force, and this can be expressed
as a change in momentum with time:

𝑑(𝑚𝑣)
𝐹= (7.1)
𝑑𝑡

The force that generates the flow (velocity) of the broth in a bioreactor
is introduced (i) by the mechanical stirrer system in a STR, or more
strictly by the motor outside the reactor, and/or (ii) by the compressed
air in the case of the airlift reactor. Instead of “force” we consider in
these biotechnology applications the “power (P)”, i.e. the rate at which
88 7 Agitation

work (force times distance) is applied to the system. The unit of power
is [W or Nm/s] and in a bioreactor the power is usually expressed in
kW. Often, however, it is more meaningful to talk about the power per
reactor volume, P/V, particularly if different reactors are to be compared.
The power drawn by most mechanically stirred reactors is 1-5 kW, and
sometimes it is as high as 10 kW. The value depends on the size of the
reactor; large reactors cannot be supplied with the high amounts that can
be used on a small scale. Consequently, there is thus a risk that a large
reactor is not well stirred.
The power input needed for a chosen mechanical agitator depends, in
order of importance, on:

I The rate of rotation, rpm


I The aeration and gas recirculation rate (a lot of air requires less
power)
I The bulk flow pattern

When we want to set up a balance equation we remind ourselves that


momentum, like mass and heat, is a conserved quantity which means
that in a closed system, momentum cannot be lost and the total quantity
stays the same. The law of conservation implies that it is possible to
make a balance over any system because what goes in must come out,
provided that the system does not (i) accumulate or (ii) consume or
produce momentum, mass or energy. Taking this into account, a general
balance can be written over the reactor where everything going in and
all that is produced must be equal to what goes out plus what might be
accumulated i.e.:

𝐼𝑁 + 𝑃𝑅𝑂𝐷 = 𝑂𝑈𝑇 + 𝐴𝐶𝐶

where ACC represents the accumulation and PROD the production. We


may use this balance either for momentum, heat or any component (mass)
of the system. If the system instead consumes a component (CONS),
rather than produces it, the balance is instead expressed as:

𝐼𝑁 − 𝐶𝑂𝑁𝑆 = 𝑂𝑈𝑇 + 𝐴𝐶𝐶

The word “system” needs some further explanation. In general, we use


the word “system” for an environment that is closed. This means that it
should be possible to limit a certain area or volume, sometimes called
the “control volume”, over which we make this balance. Figure 7.1 shows
a bioreactor where liquid (F) and gas (Q) flow are entering and exiting
the system, and where the system boundary is marked by the dotted
red line. Obviously, a bioreactor is a good example of a natural control
volume.
In the case of momentum transfer, we are however interested in the
liquid mixing inside the bioreactor and we therefore make a balance only
over a liquid element i.e. over a certain liquid volume rather than over
Figure 7.1: The bioreactor as an example of a the whole vessel. A balance over this control volume is an example of
control volume over which we can make the a “macroscopic balance” where the transport is driven by what we call
balance equations. The figure shows that both
the two liquid (F) and the two gas (Q) streams “convective” forces, which literary means that dissolved components are
pass the control boundary transported along with the solvent i.e. the medium.
7.2 Macroscopic momentum balance describing the macromixing 89

Superimposed on the macroscopic transport of liquid we have also forces resulting in a molecular level
transport which takes place by diffusion. For this we can make a “molecular transport balance” which is
also a useful and important concept in connection to the mixing as we shall see later on in this chapter.
In principle, we can thus make a balance over any surface, hereby balancing fluxes from the very small to the
very large. We can for example make a balance of fluxes as small as those inside a cell, over a whole cell,
over the bioreactor, or even over a whole industrial factory. An example of the latter is when we follow the
streams of substrates from birth (origin of the source) to cradle (products or waste) in what we call a Life
Cycle Analysis to better understand the sustainable properties of a product.
The use of balance equations is really one of the most important tools used by chemical and biochemical
engineers to understand the efficiency of a system. The concept is further used e.g. to determine unknown
fluxes and concentrations and gives us a tool to set up equations needed for a mathematical simulation of
processes. These can then be used to predict the process outcome, to understand the influence of a certain
variable, to plan the process and much more.
By convention, we generally write the balance equation in the following form which we shall use hereafter:

𝐴𝐶𝐶 = 𝐼𝑁 − 𝑂𝑈𝑇 + 𝑃𝑅𝑂𝐷(−𝐶𝑂𝑁𝑆)

7.2 Macroscopic momentum balance describing the macromixing

Let us now consider the macroscopic momentum balance over a liquid element and the forces acting upon
it. This balance is rarely used for design of commercial processes, but we write it up since it is useful for
comparison with the macroscopic mass balance which is later on used in chapters 8 and 9.
The driving force for momentum transfer is the existence of velocity gradients which develop between areas
of high and low energy input. The momentum balance is made over a liquid element in all three dimensions
i.e. in the x, y and z directions. The balance in the x-direction over such a liquid element of constant volume,
considering only the liquid and not the gas flow is given by:

𝑑(𝑚𝑣)𝑥 𝐹𝑖𝑛 𝐹𝑜𝑢𝑡 X


= 𝑚𝑣 𝑥,𝑖𝑛 − 𝑚𝑣 𝑥,𝑜𝑢𝑡 + 𝐹𝑥 (7.2)
𝑑𝑡 𝑉 𝑉

where 𝐹 , i.e. the “production” term, is a summary of the different forces that can act upon a liquid element
P
apart from those created by the ingoing and outgoing flow. This is generally made up of forces generated by
pressure, friction and gravitation. In chemistry and biotechnology, the mass, m, represents the mass of the
liquid flow and not a solid body as in classical mechanics. This means that in chemistry, the density, 𝜌, is
more often used and in biotechnology more often the concentration, “c”, i.e. mass/volume.
In general physics, the movement of a solid body is described by the equation:

𝑑(𝑚𝑣) 𝑑𝑣 𝑑𝑚
𝐹= =𝑚 +𝑣 =𝑚∗𝑎 (7.3)
𝑑𝑡 𝑑𝑡 𝑑𝑡

where we recognise that the second term of the differential can be omitted since there is no change in mass
with time, i.e. it is constant. The equation describing momentum transport in a flowing liquid, as was shown
further above, is more complex but it does not differ in principle from our earlier experience of momentum.
We leave the macroscopic balance for momentum at this stage and it will not be further treated. We realise
however that if we needed to know the rate distribution of gas and liquid in the reactor, these balances should
be solved and since there is no analytical solution this is done by use of numerical methods. By the end of this
chapter we will show how to practically determine an “overall” mixing time of the reactor, without solving of
the equations, that may be used provide a feeling for how well and at what rate our reactor is mixed.
90 7 Agitation

7.3 Molecular transport balance describing the macromixing

The momentum balance above focuses on the “convective” transport which leads to blending of the reactor
at the macro scale. In the liquid, there are also transport processes on a micro scale i.e. on a molecular level,
where momentum transport is generated by viscous friction between layers of molecules. This concept leads
to the definition of the concept of viscosity, as we shall see, which is an important parameter in cultivation.
The transport efficiency is therefore a property also of the cultivation medium and the cell morphology and
the biological properties that results in a certain viscosity thus affects the mixing, the homogeneity, the gas
dispersion and also the oxygen transfer. It is thus a good idea to learn more about the generation of the
concept of viscosity.
For derivation of this concept, let us again consider the stirred tank reactor. Where do we find a relevant
control area for the balance equation? The part of the bioreactor with the highest energy dissipation and
consequently the highest liquid and gas rates are close to the impeller. The fact that the impeller moves and
that the reactor wall is stationary results in the development of a velocity gradient. Between the impeller and
the wall, the movement is strongly “turbulent” in character, a concept which will be dealt with later in this
chapter. Since the rate profiles are therefore difficult to describe we do not use this “control area” for our
microscopic balance. We have to seek out a more idealised flow situation to define the concept of viscosity
and in the literature we find that the fluid flow between two plates is a better case for doing so.
In Figure 7.2 we study the development of a velocity profile in the medium which is inter-spacing two parallel
plates of infinite length. We note that from being at rest, the bottom plate is at time zero pulled by a force F
while the top plate continues to be at rest. This leads to the development of a “velocity profile” in the liquid
between the plates and this profile will become stationary with time i.e. the velocity profile is stable.

Figure 7.2: Development of a velocity profile between two plates of infinite length. At time=0, the medium between the plates is at rest. At the start of
the experiment the lower plate is pulled to the right in the figure with a constant force=F and a velocity profile starts to develop. The velocity is highest
closest to the moving plate and zero at the position of the plate in rest. At steady state, the profile is fully developed (bottom graph) and the velocity in
x-direction is seen as a function of the distance between the plates i.e. in the y-direction. The velocity profile, Vx=f(y), is described as a set of arrows of
different length and where each arrow represents the size and direction of the velocity at a particular point. The movement of the fluid takes place in the
x-direction i.e. in the direction of the applied force while the momentum is transferred in the y-direction, i.e. from high to low velocity

Since the upper plate is stationary, the liquid velocity closest to this wall is zero and the highest velocity is
close to the moving plate. In between, the rate of the liquid flow gradually decreases as we approach the
stationary plate. The force on a given liquid element is called the “shear force” or “shear stress”. This shear
force, denoted 𝜏, acts upon a surface area, and the unit is thus [N/m2 or Pa]. The velocity of a liquid element
at a given point between the plates is called the “shear rate”, dvx /dy, and the unit is [s-1 ]. In other words,
this
7.3 Molecular transport balance describing the macromixing 91

describes the actual rate, v, of a liquid element flowing in the x-direction


at a distance dy from the moving plate.
The flux of momentum is directed from the moving to the stationary
plate i.e. it is transferred in the y-direction. Another way of looking at
the momentum transfer is to consider the movement of the molecules
in the liquid where the initial movement of molecules is due to that the
one of the plates is set into motion. The momentum transfer from this
plate sets the closest molecular layer into movement which influences
the next set of molecules in the adjacent layer and when this layer gains
momentum it, in turn, influences the next layer and so on all the way to
the other plate. As the stationary plate is approached, each layer moves a
little bit slower. We may also say that momentum is transferred due to
the velocity gradient.
The viscosity is then a property of the liquid (the medium) that indicates
how fast, i.e. at which shear rate, the liquid will move in relation to the
applied shear force. The equation describing the relation between the
shear force and the shear rate is:

𝑑(𝑚𝑣 𝑥 )
𝜏 = −𝑣 ∗ (7.4)
𝑑𝑦

The negative sign indicates that the movement takes place from high to
low rates. This expression is analogous to Newton´s second law, where
the applied force is equal to the flux of momentum, but the balance is
here made, not over time, but over the distance “dy”. The “mass”, m, in a
liquid system is often, and as said before, expressed by the liquid density,
𝜌, i.e. the mass per unit volume [kg/m3 ] and since it is not changing
with the distance, the d𝜌/dy term of the developed differential can be
omitted.

𝑑𝜌 𝑑𝑣 𝑥 𝑑𝑣 𝑥 𝑑𝑣 𝑥
 
𝜏 = −𝑣 𝑣 𝑥 ∗ +𝜌 = −𝑣 ∗ 𝜌 = −𝜂 (7.5)
𝑑𝑦 𝑑𝑦 𝑑𝑦 𝑑𝑦

The proportionality constant is the viscosity, and “v” is called the kine-
matic viscosity [m2 /s], sometimes called the “diffusivity of momentum”.
A more commonly used expression is however the “dynamic viscosity“
“𝜂” with the unit [Ns/m2 or Pa*s] which includes the mass and which
allows a more direct comparison of the force and the apparent rate. In
the equation above we see that 𝜂 = 𝑣 ∗ 𝜌.
The viscosity usually has a low value in SI-units and is thus gererally
expressed in [mPa*s]. The equation above, referred to as Newton´s law
of viscosity, shows that the shear rate is linearly dependent of the shear
stress and media that obey this law are consequently called “Newtonian
media”. Water is an example of a Newtonian medium and the viscosity
is 1 mPa*s. A concentrated sugar solution, on the other hand, can have a
viscosity of over 1000 mPa*s.
The equation above can be plotted for pairs of shear stresses and shear
rates (with a positive sign for the shear) and we get the principal graph Figure 7.3: The viscosity, i.e. the slope given
by the equation:𝜏 = 𝜂 ∗ 𝑑𝑣/𝑑𝑡 , is constant for
in Figure 7.3. We see that if the stresses and rates are known, the slope a Newtonian medium
can then be used to derive the viscosity.
92 7 Agitation

To create this curve we thus have to apply a set of shear forces and
measure the resulting rates at each point and plot these in a diagram like
the one shown above. For this purpose we use a viscometer or rheometer.
We note here that the subject “rheology” is the study of the quality of
fluid flow, or flow of matter, which is a division of science of its own.
Rheometers are devises that principally are based on the generation of a
torque (𝜏) that is proportional to the shear rate when a certain force is
applied. The instruments range from the very simple viscometers which
may only allow the development of a single force/rate couple to those
which will allow the generation of a larger set of points that can be used
to draw 𝜏 = 𝑓 (𝑑𝑣 𝑥 /𝑑𝑦).
This simple concept does not however apply to all media. The “Non-
Newtonian media” are media with very different properties and where
the viscosity is not a constant but changes with the applied force. It is
thus very important in the design of any device including a flowing
fluid that the medium characteristics are determined and taken into
consideration. The graph below shows a summary of the principal types
of liquid behaviour. Depending on how the shear stress varies with the
shear rate the fluids are characterised as pseudoplastic, dilatant and
Bingham plastic. These different types of flow behaviour are shown in
Figure 7.4.
The Newtonian liquid is shown in the centre of the graph as the straight
line depicted by Newton’s law. The “Bingham fluid” at the top of the
graph is a special case of a Newtonian fluid in that it shows a linear
relationship between applied force and resulting rate but only after a
certain force, a yield stress, is applied. A certain force is thus needed
even to set this particular medium into motion.
The “pseudoplastic” and “dilatant” media are true non-Newtonian
media. For pseudoplastic media the viscosity is high at low shear rates
Figure 7.4: Relation between the applied shear but decreases with increasing velocity (the slope of the curve decreses)
force and the resulting shear rates of various and the applied force and such media are consequently called “shear-
fluids, both Newtonian and non-Newtoninan
thinning” media. The opposite is true of the dilatant media, which are
thus “shear-thickening”.
The behaviour of non-Newtonian media is generally described by the
“Power law equation”:

𝑛
𝑑𝑣 𝑥

𝜏𝑦𝑥 =𝐾∗ (7.6)
𝑑𝑦

where the power “n”, is the “flow behaviour index”, which is >1 for
dilatant media and <1 for pseudoplastic media. K is called the con-
sistency index, [Nsn /m2 ]. Several microbial cultivations, in particular
where Fungi are used, are non-Newtonian and often show pseudoplastic
characteristics. Since the liquid flow rate is very modest in large reactors
in spite of the applied power, the viscosity is therefore very high.
To get the corresponding expression for the Bingham fluid, we add a
term for the power needed to initially move this liquid, the yield stress
𝜏0 , and the equation consequently becomes:

𝑑𝑣 𝑥
𝜏𝑦𝑥 = 𝐾 ∗ + 𝜏0 (7.7)
𝑑𝑦
7.3 Molecular transport balance describing the macromixing 93

Some paints are characterised by this behaviour and such paints are
generally in a gel-form which only starts to flow after a certain yield
stress is applied. Other common media which are Bingham plastic are
toothpaste and mayonnaise. To measure the viscosity of a non-Newtonian
medium is never easy.
Sometimes we might however only need the viscosity in one single point
since this is were we will operate the cultivation at all times. It is here
that the term “apparent viscosity” or 𝜇𝛼 is useful to get an appreciation
of the viscosity at such a point. It should be noted that this is valid only
for one single shear stress/shear rate couple in the rheogram if the liquid
is non-Newtonian. This estimate is described by the equation

𝑑𝑣
𝜏 = 𝜇𝛼 (7.8)
𝑑𝑦

where the slope of a straight line from the origin directly to the point of
interest gives the apparent viscosity.
What causes a non-Newtonian behaviour in cultivation? The first cause
might be the product produced during a specific bioprocess and a
common example is the microbial production polymers that are secreted
to the medium. Figure 7.5 shows the rheogram of the polysaccharide
xanthan gum.
The heteropolymer xanthan is produced by the prokaryote Xanthomonas
campesteris and the molecular weight is 1.2 − 2 ∗ 106 Da. During the
cultivation, there is a large increase in viscosity and a special stirrer
configuration is required to maintain a sufficiently high oxygen transfer
rate. The pseudoplastic character of xanthan is a desired property in many
products and is used as an additive since the viscosity decreases with
increasing shear. Xanthan is thus used as a thickener in processed foods
e.g. in salad dressings where the customer appreciates the appearance of
a thick dressing but wants it to be efficiently mixed in a salad. Several Figure 7.5: Viscosity as a function of shear
rate at different xanthan concentrations. Here,
examples of the same product character are found in other microbial viscosity decreases with increasing shear rate
polymer production. One example is the production of hyaluronic acid, i.e. this is an example of a pseudoplastic broth.
As the xanthan concentration increases, the vis-
produced by Streptococci and used e.g. in cosmetics.
cosity dramatically rises (note the logarithmic
axes)
A second typical high-viscosity process is the one in which the cell
growth form causes the particular flow characteristics. This is the case in
processes based on cells forming mycelia as in the cultivation of the mold
Penicillium chrysogenum, which is used in the production of penicillin.
The process for citric acid production is another well-known example,
where the mold Aspergillus niger is used. Both have pseudoplastic, i.e.
non-Newtonian, character and are thus shear-thinning. There are also
examples of sheer-thickening media where the cultivation contains high
concentrations of yeast and some bacteria. The mechanism is in this case
that when the cells start to move, they collide and obstruct the flow.
A third situation which might lead to an increase in viscosity occurs
when cells die and dissolve. This leads to the release of macromolecules
from the cells and at high cell density considerable amounts of e.g. DNA
are present in the medium, which has been shown to affect the oxygen
transfer capacity. It is the values of K, n and 𝜏0 , in the Power Law equation,
which relate to the concentration (density) of these viscosity-creating
elements.
94 7 Agitation

We have studied the momentum balances in this chapter and have indi-
cated that also mass and energy balances can be made on both the macro-
and micro-scales. The complete set of balances in three dimensions are
called the Navier-Stokes equations and since the complete set becomes
very complex, numerical methods are used to solve them. Nowadays it is
possible to buy software packages for solving of many of the complex
problems in bioreactors. This is a research subject of it own called CFD,
Figure 7.6: CFD simulation of the flow Computational Fluid Dynamics.
through a pipe with the inlet and outlet on
the same side, but at opposite ends. Arrows An example of the solution to such a model by CFD is shown in Figure
show velocities and the colour shows the con-
7.6 where the arrows indicate the velocity gradients in the fluid showing
centration. Red is a high concentration and
dark blue a low concentration. The CFD image their magnitude and direction. Such solutions contain obviously a lot of
was here created by the software “Fluent” valuable information for process design.

7.4 Mixing devices

In the chapter on bioreactors it was mentioned that there are two major
ways of stirring of a liquid; viz. through mechanical devices (STR) and
through dispersion of large quantities of compressed air into the liquid.
We will in this chapter consider the former.

Agitator types

Two radically different types of stirrers are used, the flat blade Rushton
turbine and the standard (marine) propeller but a range of variants and
combinations of these are also available.
The marine propeller is often used when high shear forces have detri-
mental effects on the chosen cell type since these forces are minimised by
this design. They are thus more or less the golden standard in animal cell
cultivation since these cells are more shear sensitive than microorganisms.
Figure 7.7: The two basic types of mechanical The flow created by a marine impeller is mainly directed straight up- or
stirrers. Left: the flat blade Rushton turbine (6 downwards, i.e. it leads to an axial pumping flow. This results in that the
blade) and right: the marine propeller (3 blade)
radial flow is not promoted, which is a drawback that has to be accepted
in order not to destroy the cells. Several variants of this basic pumping
design are available.
The flat blade Rushton turbine on the other hand, is a radially and
tangentially pumping device. The high shear forces created combine
good mixing with the ability to shear the gas bubbles into smaller
structures which is very important for good aeration as we shall discuss
later on. These two different impeller types create radically different
mixing patterns in the bioreactor and the principal outline is shown in
Figure 7.8. In reality, mixing in most reactors is chaotic and patterns like
the ones in the figure are based on average rates over time. Simulations
of the mixing are often performed using the commercial CFD software
as shown in the figure.
7.4 Mixing devices 95

Figure 7.8: Mixing patterns (B,D in bioreactors using different impeller designs as shown by CFD simulation. Left: the average pattern generated
by the Rushton impellers (A) and right: the pattern from pitched blade turbines which are variants of a propeller (C). The figure is presented with
permission from Prof. A Lübbert, Univ. of Halle

Simulation can also be used to show the mixing of two liquids at intermittent times. Figure 7.9 shows the
mixing of two liquids (red and blue) using a downward pumping impeller and the eventual (near) equilibrium
concentration.

Figure 7.9: Time series of mixing of two liquids, one red and one blue, in 20 s. A downward pumping impeller is used

Many different intermediate agitator configurations between the two main types exist, as we have indicated
from the start, and companies sell their own designs. Commonly used axial impellers are shown in Figure
7.10, i.e. variants of the propeller.

Figure 7.10: Commonly used axially pumping impellers. From left to right: a pitched blade 45° up pumping impeller, the MaxFlo W down pumping
impeller, the Lightning A315 down pumping impeller and the Lightning A600 down pumping impeller

A Rushton type impeller may be preferred when good oxygen transfer is critical and since many cultivations
are aerobic, this is a frequently used design. Also here retailers have their own versions as shown in Figure
7.11.

Figure 7.11: Commonly used radially and tangentially pumping Rushton impellers. From left to right: the Chemineer BT6 hydrofoil, the Chemieer
CD-6 hydrofoil, the paddle and the Rushton curved impeller
96 7 Agitation

With a Rushton impeller, the movement of the impeller perpendicular to


the reactor shaft creates a large under-pressure behind the blade (see the
simulation at the top of Figure 7.12). This has several consequences. One
is that gas bubbles are attracted to this region and bubbles are “merged”
together. This capture of the gas phase behind the blade creates large
cavities which are discharged at the upper and lower edges of a Rushton
impeller (see the bottom of Figure 7.12). At a fixed stirrer speed the power
drawn by the impeller will therefore be reduced due to the entrapped air
- which is positive, but since the bubbles are larger, the oxygen transfer
is reduced. Nevertheless, the Rushton agitator has still very high gas
dispersion properties. When the air eventually leaves this zone in large
packages, it might however create an unstable operation situation due
to the sudden variation in the power drawn by the impeller. This is not
desired and to reduce this behaviour, and reduce the cavities, the curved
blade Rushton impellers were designed with a concave pipe segment
called a “hollow” blade instead of a flat one. Examples are the Chemineer
CD6 (Figure 7.11) and the Swedish “Scaba” agitator (not shown). The
risk of cavitation can also be counteracted by the use of more than 6
blades. It should be noted that a high viscosity leads to increased risk for
Figure 7.12: Top: Pressure gradients behind cavitation.
the impeller blade as shown by CDF simula-
tion. The arrow indicates the rotation of the
impeller. Bottom: trailing vortex created from
gas entrapment behind the impeller. As the cav- 7.5 Turbulence
ities are very long they may become unstable
and disengage as a large bubble from the im-
peller. The figure is presented with permission The reason for the good mixing capacity of the Rushton turbine is that
from Prof. A Lübbert, Univ. of Halle
it creates a high degree of “turbulent” flow or “turbulence”. Turbulent
flow is an unsteady flow, characterized by the formation of eddies in the
liquid, which are described by their distribution in size and intensity. The
opposite of turbulent flow is “laminar” flow, which is a steady flow in one
direction only. Turbulence is important and desirable, if it can be realized
and afforded, since it greatly enhances the transport properties.
Dimensionless numbers are quantities that are frequently used both
in chemical and biochemical engineering for purposes of design and
understanding of process performance. These numbers are principally
comparisons of important forces and parameter characteristics in the
processes. The first example is the Reynolds number, NRe , a number
which gives an indication of whether turbulent properties are present
or not. A high number means a high degree of turbulence. The number
Figure 7.13: Laminar and turbulent flow char-
is calculated as the ratio of the forces of inertia (Swedish: tröghet), i.e.
acteristics as shown by the addition of a tracer the fluid resistance to movement (Newton’s first law) in relation to the
(blue dye). In the laminar region, the flow fol- apparent viscous forces. The Reynolds number for a bioreactor system is
lows the streamlines (mean flow) but in the
turbulent area the velocity fluctuates in all di-
principally calculated as:
rections. Principally, large eddies lead to lateral
flow across the streamlines and small eddies
(𝑚 ∗ 𝑣) ∗ 𝑐 ℎ𝑎𝑟 𝑎𝑐𝑡𝑒𝑟𝑖𝑠𝑡𝑖𝑐 𝑙𝑒𝑛 𝑔𝑡 ℎ
lead to a spreading of the dye by diffusion 𝑁𝑅𝑒 = (7.9)
𝜂

As before, the mass unit used for fluids is generally the density. The rate
(v) and length that are characteristic for bioreactors are the tip speed
(𝑣 = 𝜋𝑁 𝐷𝑖 ) and the impeller diameter (Di ), respectively. If we
7.5 Turbulence 97

Figure 7.14: Rushton diagram relating the Power and the Raynolds dimensionless numbers. Impellers: (a) flat blade Rushton turbine, (b) paddle and
(c) marine propeller. Valid for Dt /Di =HL /Dt =3

incorporate the particular parameters for bioreactors, Reynolds number is thus expressed as:

𝜌 ∗ 𝑁 ∗ 𝐷𝑖2
𝑁𝑅𝑒 = (7.10)
𝜂

If NRe > 10 000, the flow is considered to be fully turbulent. From this expression it is evident that the greater
the momentum, the larger is the Reynolds number and the higher the turbulence. The Reynolds number
decreases however with increasing viscosity.
The Power number, NP , is the second dimensionless number we explore which is expressing the power input
per unit mass (here again: 𝜌). The power number is thus essentially a quotient between the external and the
inertial forces and the higher the number the lower is the impact of the medium and cells on the mixing
efficiency.
The parameter, Po [W or Nm/s or kgm2 /s3 ], denotes the ungassed power (i.e. no air is introduced to the
vessel).

𝑃𝑜
𝑁𝑃 = ∗ (𝑔) (7.11)
𝑁3 ∗ 𝐷𝑖5 ∗ 𝜌

From power measurements at different stirring speeds (N) in a particular reactor setup the logarithm of the
Power number is often plotted against the Reynolds number for various types of impeller designs. Such a plot
is sometimes called a Reynolds diagram and can be used to introduce and pinpoint the exact locations of
different “flow regimes” in the vessel at certain stirrer speeds. Such regimes are either laminar, intermediate
or turbulent.
If we can calculate our Reynolds number we may use of the Reynolds diagram to estimate the Power number
which, in turn, can be used to calculate the power drawn. The parameter “g”, in brackets in the expression
above, is the acceleration of gravity [m/s2 ] that has to be used to derive the power number if the power unit
is horsepower [HP or kgm/s]. Many diagrams and calculations, in particular in the American literature, use
this unit.
The power (P) supplied to the agitator from the motor is to some extent lost as heat e.g. in seals and bearings.
The rest is lost mostly in the area immediately surrounding the impeller, where as much as 70% of the energy
can be dissipated, and in the remainder of the reactor little momentum is gained. This “energy dissipation
98 7 Agitation

rate”, 𝜀𝑇 , is sometimes used to describe the efficiency of the turbine mixing ability and is independent of
scale but depending on the density of the broth.

𝑃 1 𝑊 𝑚2
 
𝜀𝑇 = ∗ 𝑜𝑟 3 (7.12)
𝑉 𝜌 𝑘𝑔 𝑠

The power drawn by an impeller at a certain stirrer speed is greatly reduced when air is supplied, since
the air lowers the medium density. Some stirrers can in fact not be started until air is introduced. We will
introduce some useful expressions later in the section on Aeration.
In large bioreactors, a number of impellers are positioned above each other on an axial shaft. A combination of
both Rushton and propeller-type impellers is then sometimes used. This allows both good oxygenation (from
the Rushton impeller) and a thorough mixing in all three dimensions of the reactor through the combination
of the axial pumping from the propeller and radial/tangential mixing from the Rushton turbines.
The distance between the impellers is an important design parameter. Impellers have to be spaced so that
a high liquid outflow from one impeller reaches the impeller immediately below or above, so that a high
degree of intermixing between impellers is created. If this is not achieved, each impeller will act as an
“isolated” bioreactor and the fluid around it will be more or less stagnant in this area. The distance between
the impellers is usually 1.5-2 impeller diameters, but if the broth is highly viscous this can be lowered as far
down as to 0.5 diameters

7.6 Measurements of mixing

Measurements of mixing One way to get an image of the mixing pattern of a smaller transparent reactor is to
use decolorisation experiments. An example of this technique is shown in Figure 7.15 (top view) with the
corresponding CFD simulation results (bottom view) for a reactor with three pairs of Rushton impellers. The
characteristic patterns are clearly revealed by both techniques which are strikingly similar and shows the
strength of the particular CFD model. It should be kept in mind that decolorisation experiments is of course
not possible to perform in large commercial reactors since these are not transparent.

Figure 7.15: Comparison of CFD simulation (lower row) with experimental results (upper row) from a decolourisation experiment. The liquid is
coloured with iodine at start whereafter thiosulphate is added. The formation of iodide in the presence of starch removes the dark colour. The regions
which are poorly mixed are shown by the sustained dark colour. The reactor is of 400L equipped with three pairs of standard Rushton impellers operated
at a P/V=90 [W/m3] which is typical for animal cell cultivation. The time to homogeneous mixing is here approx. 53s. With permission from Prof. A.
Lübbert, Univ. of Halle

When we talk about measuring the mixing, we are thus mainly concerned with the liquid convective flow i.e.
the liquid bulk mixing or the macromixing.
In large commercial reactors we cannot use decolorisation experiments to estimate the mixing. To estimate
the performance in such reactors we may then conduct a series of so called “stimuli-response” measurements
to determine the “mixing time”. The definition of mixing time is the time it takes to achieve homogeneity
7.6 Measurements of mixing 99

in the reactor after we have made an external addition of a specific


component to the reactor. This implies that the larger the reactor the
longer is the mixing time. The mixing time is in the order of seconds in
small bench-scale reactors but may be several minutes in a large reactor.
The mixing time is abbreviated in many ways in different literature
such as: 𝑡 𝑚 , 𝑡𝜃 or just 𝜃 . The stimuli-response technique builds on: (i)
introduction of a selected tracer in one point of the reactor, (ii) the tracer
distribution by the generally depicted circulation pattern of the chosen
mixing device and (iii) the monitoring of the tracer appearance in another
part of the reactor.
The choice of the different positions for the addition and measurement
of the tracer will influence the eventual use and interpretation of the
data. The point of addition is often chosen as the point where we during
cultivation will add something that is of high significance to the process
e.g. the feed of a solution of highly concentrated glucose and where we
want to know if this takes a long time or not. If the mixing time is long,
we might suspect that concentration gradients of this compound might
appear.
A suggested setup is shown in Figure 7.16 where the addition is made
close to the impeller which would lead to that the addition will be
introduced in the turbulent outflow of this devise and will in theory be
moving according to the general circulation pattern shown in the figure.
This would be the most ideal point to add something form a mixing
point-of-view but in reality such points are very few and one must use
what is available. Very often additions are thus made from the top since
reactors are technically more easily fitted with such points in the separate
lid rather than in the large cylinder shape of the reactor where we have
to penetrate the double wall of the jacket which is preferably avoided.
A common stimuli-response technique is to introduce a pulse of acid or
alkali and to measure the response by a pH electrode. The acid or alkali
is added at a desired position in the reactor called the addition point. The
response that will be detected by the pH electrode is positioned in another
critical part (measuring point) well away from the injection site. In a Figure 7.16: Top: Pressure gradients behind
large reactor, the acid or alkali is not immediately mixed but will move the impeller blade as shown by CDF simula-
tion. The arrow indicates the rotation of the
according to the mixing pattern and the signal from the pH electrode is
impeller. Bottom: trailing vortex created from
supposed to show the characteristics of this mixing. The addition will gas entrapment behind the impeller. As the cav-
eventually, when mixed to homogeneity, lead to an overall change in ities are very long they may become unstable
and disengage as a large bubble from the im-
the pH. Other additions/responses might be e.g. salt addition/redox
peller. The figure is presented with permission
measurement and hot water addition/temperature measurement. from Prof. A Lübbert, Univ. of Halle

It is very important that the tracer pulse is of short duration but of a


high magnitude (e.g. a very high concentration in the case of a chemical).
Furthermore, the pulse itself should not be added in such a way that
momentum is introduced with the pulse which can result in a large
error.
The signal from a pH electrode at the measuring position may change
with time as shown by the principle outline of the response curve in
Figure 7.17. Before the start of the experiment, to the left in the graph, the
pH electrode detects the initial stable pH. At time=0, a short distinctive
pulse of alkali is added in some part of the reactor and at the time
represented by “1” in the graph, this alkali pulse has reached the position
of the electrode and the electrode reacts by showing an increase in
100 7 Agitation

Figure 7.17: Mixing time measurement by a stimuli-response experiment. A pulse of alkali is introduced to the reactor at time zero and the response
is sometime after this addition detected by a pH electrode. The diagram shows the pH electrode response curve

the pH value. We might think of the alkali pulse as a fairly concentrated “package” added to the reactor
which is not immediately dissolved but initially kept more or less intact. As this “package” continues to
circulate in the reactor the mixing and diffusion will dissolve this package and the pH at the position of the
electrode drops with time as compared to the initial response. Thus, after the added alkali has made one
repeated circulation in the reactor it is diluted but is still visible (points 1, 2 and 3). This is repeated until
the liquid in the reactor is well mixed and a new steady state pH is reached. Usually the concept of 90% or
95% mixing time is used. This is the time taken for the pH signal to come within ± 10% to ± 5% of the final
value, respectively. In the figure, the t90 occurs during the third pulse, as shown by the fact that the signal is
entering the region between the dotted red lines. The reason to the use of a lower value than 100% is that in
large reactors it is often not meaningful, or even possible, to wait for mixing to complete homogeneity.
The time between two “pulses” detected by the pH electrode is consequently called the circulation time,
tc , and this indicates a “full circulation” in the reactor before the trace has returned to the sensor (the time
between two detection points in the curve above). There is an empirical relationship between these two
mixing-related times which shows that it generally expected to take four circulations before a 90% mixing is
reached. In the former figure this takes approximately three circulations. This equation can of course not be
used to calculate the mixing time since the circulation time is even more difficult to measure.

𝑡 𝑚 90 = 4 ∗ 𝑡 𝑐 (7.13)

It should be noted that this concept is of course an oversimplification in that it proposes a stereotyped
circulation pattern for all bioreactors. However, this average pattern might still be very useful and is the only
way to practically get some understanding of the degree of reactor mixing.

Figure 7.18: Mixing time measurements in a 12 m3 stirred tank reactor with down-pumping three impellers. With permission by Prof. A. Lübbert,
Univ. of Halle
7.6 Measurements of mixing 101

In Figure 7.18, data is shown from a real mixing time experiment in a 12 m3 unaerated animal cell culture
reactor operated at 50 W/m3 . The reactor is equipped with three down pumping pitched blade impellers.
The addition of a salt pulse is done at the top of the liquid surface, a so-called top-to-bottom mixing time
experiment. Three conductivity sensors positioned in the top, opposite to the addition point, in the middle
and at a bottom position of the reactor were used to detect the response. The probes were positioned close to
the reactor wall in-between the baffles. The 95% mixing time is estimated to 96 seconds, which is the time
when the all probes are within ± 5% of the final value as is shown in Figure 7.18.
The simplified principle of mixing that stimuli-response experiments are based on, can be distinguished
from the “real” reactor measurements above and we see that the pattern varies with respect to which of the
three probes we study. Here only the middle and upper probes show an oscillating response pattern, which
might be due to the expected and repeated circulation flow in the reactor and/or from establishment of
compartments around the impellers. It is also obvious that the bottom probe reacts first in spite of being
at the most distant position from the addition point. This is typical for the axial down-pumping impeller
setup that rapidly pumps the tracer downwards close to the shaft. Eventually, all probes end up at a common
steady state value that shows the mixing time.

Figure 7.19: Mixing time measurements in a 400 L stirred tank reactor. Dependency of the addition point of a compound, in relation to a top addition,
on the resulting mixing time. Curtsey of Prof. A Lübbert, Univ. of Halle

From the data in Figure 7.18 it is of course realised that it is important to choose a good position for additions
to the reactor if we want a fast mixing of an added component. Figure 7.19 below shows how the mixing time
is reduced depending on where the trace is added compared to addition at the surface for the same reactor
setup as above.
The reactor is shown to the left of the figure and the different colours indicate the four additions points.
Additions in the middle position seems here to be the preferable since the mixing time reduction is the
highest i.e. 34%.
In reality we may expect that bioreactors will not operate according to the general mixing patterns in particular
when we are using impellers in stirred tank reactors which is probably due to the large degree of random
motion in these. We might instead have to accept that we get compartmentalisation in large tanks where the
intermixing between the compartments might be very restricted or that the liquid, in the other extreme, just
moves along the sides. However, when we introduce a large degree of aeration into the vessel, this will also
have a large impact on the performance.
102 7 Agitation

Figure 7.20: CFD simulated trajectories of liquid tracer particles in the surrounding of the different impellers showing the difficulty to escape the
specific impeller regions. With permission from Prof. A. Lübbert, Univ. of Halle
Cultivation Concepts 8
There are different techniques, or technical concepts, available for cultiva- 8.1 The general mass balance con-
tion and which are used to steer the performance to a desired end point cept . . . . . . . . . . . . . . . . . . 103
in cultvation. So far only one cultivation concept has been discussed: The balance of substrate over the
reactor . . . . . . . . . . . . . . . . 105
the batch process, where all nutrients are added in the beginning of the
The balance of cells and product
cultivation and from which we get the growth curves discussed in chapter
over the bioreactor . . . . . . . . 105
4. This procedure does not permit a high degree of interference with the
8.2 Batch processing . . . . . . . 106
process as it runs and it is therefore not used very much in the industrial 8.3 Continuous cultivation . . . 107
setting. There are however three essentially different techniques available Continuous cultivation in prac-
for cultivation of living cells: tice . . . . . . . . . . . . . . . . . . 108
The general liquid side mass bal-
I Batch cultivation ance for a chemostat . . . . . . . 109
I Continuous cultivation The balance of cell mass in a
I Fed-batch cultivation chemostat . . . . . . . . . . . . . . 109
Mass balance of the limiting sub-
The two latter concepts build on the addition and sometimes also the strate . . . . . . . . . . . . . . . . . 110
removal of matter from the reactor. Since this will affect the accumulation The mass balance of other sub-
of the important compounds inside the reactor, i.e. the cell, substrate strates than the limiting . . . . . 111
and product concentrations, of which we are naturally highly interested, A plot of parameters against the
we need a method to calculate the effect of addition of such ingoing dilution rate . . . . . . . . . . . . 112
and outgoing streams. So before studying the continuous and fedbatch Reasons for restriction of the
techniques in detail, we need to know the technique of how to make a growth rate . . . . . . . . . . . . . 114
macroscopic mass balance over the bioreactor. The mass balance builds The chemostat performance 114
on the same balance equation principle that was introduced in the Effects of the saturation con-
stant . . . . . . . . . . . . . . . . . 115
preceding chapter for the momentum balance.
The self-regulating metabolic
control of the chemostat . . . . . 116
Applications of the chemostat117
8.1 The general mass balance concept Continuous cultivation with re-
circulation . . . . . . . . . . . . . 119
8.4 Fedbatch cultivation . . . . . 119
Consider the figure of the bioreactor of the volume V (Figure 8.1), where
Mass balances in fedbatch culti-
a randomly selected component “y” [kg/L] is produced or consumed at
vation . . . . . . . . . . . . . . . . 121
the volumetric rate, ry [kg/L,h]. This component, y, can e.g. be the cell The balance for cell mass and
mass or the substrate. We note that such a component can, under certain product in fedbatch cultivation 121
circumstances, be introduced with the liquid (Fin ) and/or by the gas flow The mass balance for substrate in
(Qin ). It may also leave the reactor with any of the outgoing streams of fedbatch cultivation . . . . . . . 122
liquid or gas, Fout [L/h] and Qout [L/h], respectively. The substrate feed . . . . . . 122
The maximum cell density . 127
It is here the solution to the expression of the differential dy/dt (the accu-
mulation term) that we want to determine since solving this differential
equation will give the actual concentration of the important parameters
x, s and P. We note that we make the balance over the reactor since it is
over this boundary (control volume) that we transfer “substances”, or
“mass”. This will apply to all components that we might add or remove.
The balances for oxygen and carbon dioxide we will however come back
to in the next chapter on aeration.
We may however initially start by noting that we can make a balance
on the flows as such. The balance for the liquid flow, provided that no
104 8 Cultivation Concepts

liquid is accumulated inside the reactor, is very simple: the input flow
equals the output flow.

X X
𝐹𝑖𝑛 = 𝐹𝑜𝑢𝑡 (8.1)

We recognise that in the Figure 8.1, there is just one ingoing and one
outgoing flow which of course makes things very simple but mirrors not
often the reality.
The corresponding gas flow rate balance should take into account that
the biological reaction produces gas (i.e. carbon dioxide) which might
not always be proportional to the consumption of oxygen. To solve this
Figure 8.1: Start-up for a mass balance over we have to go over to a component balance in the gas and in this case it is
a bioreactor for any component, here denoted
by “y [g/L]” showing the liquid flow, F, in and common to make the balance on nitrogen (Q*cN2 ). The trick of doing so is
out of the reactor as well as the corresponding that the nitrogen gas is inert in the sense that it does not participate in any
gas flow, Q. V is the liquid volume, dy/dt the reaction inside the reactor. The nitrogen balance is thus very simple and
rate of accumulation and 𝑟 𝑦 the volumetric
reaction rate for the component “y”. NOTE! can be written as follows, where IN=OUT (no production or consumption
The concentration of “y” is the same in the terms are present regarding nitrogen).
reactor as in the outgoing flow! The broken red
line shows the control volume over which the
mass balance is made
𝑄 𝑖𝑛 (100 − 𝑐 𝑂2 ,𝑖𝑛 − 𝑐 𝐶𝑂2 ,𝑖𝑛 − 𝑐 𝑤,𝑖𝑛 ) = 𝑄 𝑜𝑢𝑡 (100 − 𝑐 𝑂2 ,𝑜𝑢𝑡 − 𝑐 𝐶𝑂2 ,𝑜𝑢𝑡 − 𝑐 𝑤,𝑜𝑢𝑡 )
(8.2)
This balance is made on the assumption that the gas contains only
nitrogen, carbon dioxide and water vapour (w). The water vapour and
carbon dioxide contents in the inlet gas are often close to zero and the
oxygen concentration in air is well known and since the ingoing gas
flow rate is normally measured all components on the left side of the
equation are accounted for. The outlet gas flow is our unknown. The
concentrations of oxygen and carbon dioxide in the output gas are also
often measured and we note that very little water vapour leaves the
reactor since we use a condenser to capture liquid in the outlet gas. This
means that all the components are known except for Qout. Note also that
the balance is made in terms of volume% and not g/L or mmol/L. The
use of volume% is very common for gases since this is what we mostly
measure. The approximation of the outgoing flow can then be written
as:

79.05
𝑄 𝑜𝑢𝑡 = 𝑄 𝑖𝑛 (8.3)
100 − 𝑐 𝑂2 ,𝑜𝑢𝑡 − 𝑐 𝐶𝑂2 ,𝑜𝑢𝑡

where 79.05% is the common fraction of the nitrogen in the ingoing


gas.
To continue with mass balances of cells, substrates and products, we note
that these may be transferred by the flow of liquid (F) or the gas (G) or
both. We will make the balances over the individual flux of a specific
component, which means that we look at the flow of mass i.e. F*y with
the unit [kg/h]. To make such a mass balance of a component, we need to
take into account that these are often being consumed (such as substrates)
and may also be produced and accumulated at certain rates (which is
hopefully the case with regard to the product!). We return first to the
schematic figure of the bioreactor. To make the mass balance, we use
again the law of conservation of mass and we derive what we call “the
8.1 The general mass balance concept 105

general mass balance equation” which is valid for any component. Since we are concerned only with the
balances on substrate (excluding oxygen), product and cells, we omit for the present purpose the gas flow
where these are rarely present. The general balance, as shown before, will build on the correlation:

𝐴𝐶𝐶 = 𝐼𝑁 − 𝑂𝑈𝑇 + 𝑃𝑅𝑂𝐷/(−𝐶𝑂𝑁𝑆)

For any component, “y” [kg/L], that is not transferred between phases, whether it be substrate, product or
cells, the general mass balance equation will be:

Definition 8.1.1 The general mass balance equation

𝑑(𝑦 𝑜𝑢𝑡 ∗ 𝑉)
= 𝐹𝑖𝑛 ∗ 𝑦 𝑖𝑛 − 𝐹𝑜𝑢𝑡 ∗ 𝑦 𝑜𝑢𝑡 ± 𝑟 𝑦 ∗ 𝑉 (8.4)
𝑑𝑡

We note that the volume falls inside the differential since it may change during the process. It is good to stop
the reading here and take some time to make an analysis of the units of each term!

The balance of substrate over the reactor

Why are we making the balance on substrate? Well, we want to know how this concentration in the reactor
changes with time and certainly the time when it is finished! This is the same as solving for “cs ” in dcs /dt
where “y” is replaced by “cs ”. Let us consider the different terms of the general equation above and now
with regard to the substrate glucose (s).
Both the input and output to the reactor may contain glucose, so both these terms have to be included.
Inside the reactor, glucose is consumed due to cell growth and perhaps there will also be an accumulation of
glucose if it cannot be used at the rate supplied. The general balance equation for glucose, for any cultivation
technique used, is therefore,

Definition 8.1.2 Mass balance equation for non-volatile substrates

𝑑(𝑐 𝑠,𝑜𝑢𝑡 ∗ 𝑉)
= 𝐹𝑖𝑛 ∗ 𝑐 𝑠,𝑖𝑛 − 𝐹𝑜𝑢𝑡 ∗ 𝑐 𝑠,𝑜𝑢𝑡 − 𝑟 𝑠 ∗ 𝑉 (8.5)
𝑑𝑡

The equation shows which parameters that affect the substrate concentration in the reactor with time. The
liquid volume, V, may change during the process since the input and output flows are not necessarily the
same. The volume is thus included in the differentiation. The output concentration of glucose is of course the
same as the concentration inside the bioreactor. As shown in an earlier chapter, the volumetric consumption
of glucose is given by 𝑟 𝑠 = 𝑞 𝑠 ∗ 𝑐 𝑥 and this term can be inserted into the equation instead of 𝑟 𝑠 . The sign is
negative since glucose is consumed.

The balance of cells and product over the bioreactor

The same reasoning as for the substrate can also be used for the production of cells. No cells or products are
however present in the input liquid flow. The resulting balances are thus:
106 8 Cultivation Concepts

Definition 8.1.3 Mass balance equation for cell mass

𝑑(𝑐 𝑥,𝑜𝑢𝑡 ∗ 𝑉)
= −𝐹𝑜𝑢𝑡 ∗ 𝑐 𝑥,𝑜𝑢𝑡 + 𝑟 𝑥 ∗ 𝑉 (8.6)
𝑑𝑡

Mass balance equation for non-volatile products

𝑑(𝑐 𝑝,𝑜𝑢𝑡 ∗ 𝑉)
= −𝐹𝑜𝑢𝑡 ∗ 𝑐 𝑝,𝑜𝑢𝑡 + 𝑟 𝑝 ∗ 𝑉 (8.7)
𝑑𝑡

for the product concentration. In both cases, we have of course to include a production term which is
positive.
We have now the tools to take a step further in the study of the different cultivation concepts and to investigate
these into detail. Let us however start with the common batch process that we have already talked about in
preceding chapters to get familiar of how to use the balances.

8.2 Batch processing

The main characteristic of a batch process is that all the nutrients needed for the entire process (except for
oxygen!) are added before inoculation. The process is then run until a point at which the process is shut down.
This is a point which is most carefully designed. This procedure is then repeated for the next production run
with the steps indicated in Figure 8.2. For a considerable part of the time in the factory, this type of cultivation
leads to long down-times when no product is formed which is a primary drawback with this method.

Figure 8.2: Production steps that are continuously repeated in batch processing and where the “running phase” is the only productive time

As has been shown earlier, a specific cell will in a batch process grow at the maximum rate set by the
prevailing temperature, pH and substrate characteristics. Growth at the maximum rate means however also
a maximum consumption of oxygen. The oxygen consumption rate, ro, has to be balanced by the capacity
for oxygen transfer to the medium and this has also a maximum value for each bioreactor. Because of this
limitation, and the very low solubility of oxygen in water, growth can only continue to a comparatively
low cell density in a batch process since the oxygen consumption is at its maximum. This is a productivity
drawback that cannot be tolerated in most industrial processes and this means that the batch concept is only
used when a very high substrate concentration is needed and this is true of very few processes.
When the general mass balance that we just derived is used to set-up the balances for a batch cultivation, we
observe that there is no liquid flow either into or out of the reactor, i.e. Fin = Fout = 0, and the three equations
for substrate (glucose), cell mass and product are consequently reduced to:
8.3 Continuous cultivation 107

Definition 8.2.1 Mass balance equations in a batch process

𝑑(𝑐 𝑠 ∗ 𝑉)
= −𝑟 𝑠 ∗ 𝑉 (8.8)
𝑑𝑡

𝑑(𝑐 𝑥 ∗ 𝑉)
= 𝑟𝑥 ∗ 𝑉 (8.9)
𝑑𝑡

𝑑(𝑐 𝑝 ∗ 𝑉)
= 𝑟𝑝 ∗ 𝑉 (8.10)
𝑑𝑡

The lack of any inflow and outflow means that the volume remains constant, so that the parameter V can
be taken outside the parentheses on the left-hand side and then deleted from both sides. Since 𝑟 𝑠 = 𝑞 𝑠 ∗ 𝑐 𝑥 ,
𝑟 𝑥 = 𝜇 ∗ 𝑐 𝑥 and 𝑟 𝑝 = 𝑞 𝑝 ∗ 𝑐 𝑥 , the equations can then be rewritten as:

𝑑𝑐 𝑠
= −𝑞 𝑠 ∗ 𝑐 𝑥 (8.11)
𝑑𝑡

𝑑𝑐 𝑥
= 𝜇 ∗ 𝑐𝑥 (8.12)
𝑑𝑡

𝑑𝑐 𝑝
= 𝑞𝑝 ∗ 𝑐𝑥 (8.13)
𝑑𝑡

We observe that these are the same equations as those shown in Chapter 5 when we discussed the concept of
“productivity”.

8.3 Continuous cultivation

In the discussion of batch cultivation above, we said that the oxygen consumption rate, 𝑟 𝑜 , is one of the
limiting factors for this cultivation concept since cells grow very rapidly and are thereby demanding a large
oxygen availability. So, the question is whether we can introduce a cultivation concept in which the oxygen
consumption can be restricted and whether this can be used to increase the productivity.
Our former stoichiometry considerations and the derived cell flux model have shown us that oxygen is used
for the production of energy (in turn for the making of new cells) meaning that there is a molar relationship
between oxygen consumption and cell mass production. This can be further emphasized by use of the yield
concept, here by the relation between cell growth and oxygen consumption, where 𝑌𝑥𝑜 = 𝜇/𝑞 𝑜 . This means
that if the yield is reasonably constant, a reduction in the growth rate will lead to lower oxygen consumption.
This seems good enough for our purposes but how then can the growth rate be reduced?
The answer is found if we look at the Monod equation, which shows how the growth rate develops as a
function of substrate concentration. From this we see that if the substrate concentration is reduced below the
threshold value (look back at the graph!), the growth rate will decrease and consequently; the consumption
of oxygen can be reduced.
In practice this is done by removal of a component from the medium that can limit growth – and from the
former chapters we may understand that this is most effectively done by removal of the carbon source and the
target is often glucose. If this substrate is instead fed to the reactor at a restricted rate, that we set ourselves,
the rate of growth can be controlled. This is actually the basis for both continuous and fed-batch operations. It
should perhaps be emphasised that the limiting component does not need to be glucose but may be any
108 8 Cultivation Concepts

substrate that is capable of limiting growth. In Chapter 4, we had e.g. a look at the effects on growth of
limitation of phosphorus and nitrogen.
There are some different manners of how to manage a continuous cultivation and they build on how the
limiting concentration of the selected substrate is controlled. The most commonly used continuous cultivation
concept is the “chemostat” and the manner of control of this technique is shown in this chapter into
greater detail. Another continuous cultivation process control scheme is the “nutristat” where the limiting
component (e.g. glucose) is measured in the reactor and this signal is fed back to the liquid pump which
balances the growth by increasing or decreasing the feed of new glucose. Two other continuous cultivation
control concepts have been described in the literature: the “turbidostat” and the “pH-auxostat” where the
measurement either of cell mass concentration or of the pH-controlling substance (which only under specific
conditions is proportional to the growth rate) is used to control the feed. The latter two concepts can however
only be used at the maximum growth rate and this is usually not a benefit in production.

Continuous cultivation in practice

The practical set-up of a continuous cultivation is shown in Figure 8.3. In most cases a continuous cultivation
is started after a short batch phase, which is long enough for the cells to adapt to growth in the reactor i.e.
they should preferably grow into the log phase. This is done by calculation of a batch medium with a certain
amount of the substrate which is selected to eventually be the limiting. This design will allow growth to a
certain desired cell concentration for the continuous experiment. If this substrate is depleted, e.g. if we chose
glucose, the cell would stop to grow. For the start of the continuous cultivation we will however not wait
until this point.
Instead, a complete medium is allowed to enter the bioreactor from a reservoir via a pump at a preset speed.
The pump speed sets the specific growth rate and the medium content of limiting substrate will determine
the cell concentration, as we will see later on. The upper limit of the cell concentration must be low enough to
give a growth restriction inside the bioreactor not to exceed the oxygen availability at high growth rate. This
is a basic limitation of continuous cultivation i.e it can not run at particularly high cell concentrations if not
the growth rate is kept very low. We note also that the concentration in the reactor, of the substrate which
restricts the growth, lies in the mg/L range for most cell types and substrates.

Figure 8.3: Continuous cultivation. The yellow reservoir contains the medium with the growth-restricting factor dispersed in medium. Both the inlet
and outlet liquid flows are controlled from relevant set points and will lead to that the volume is always constant

To avoid volume build-up, the inflow rate, Fin , is balanced by an equal outflow rate, Fout . Provided that these
flows are indeed balanced, the volume remains constant. Various methods are used to achieve this. In the
8.3 Continuous cultivation 109

figure, the liquid level is continuously measured and recorded by a conductivity sensor. As soon as the liquid
reaches the sensor, a signal goes to the pump, which starts to remove the overflow liquid. Another technique
is the place the reactor on a balance and in this way keep the volume constant. The inflow is calculated
according to mass balance principles and such calculations are shown in the following.
Continuous cultivation using the chemostat principle is used for many different purposes: for production,
as a tool for process development and for researchers to get deeper understanding of the cell physiology
during processing. A unique feature of the chemostat is that this may be the only way to cultivate cells with a
defined physiology for a long time, since all the parameters are at “steady state” and do not change with
time, as we shall see. If we feed new substrate to the reactor, the technique provides also the possibility to run
the cultivation “for ever” with a continuous product outtake. This will lead to a very high productivity since
we will not have any shut-down periods.

The general liquid side mass balance for a chemostat

To dig a little bit deeper into the chemostat principle let us consider the general mass balance, now for
continuous cultivation. We continue to consider only the liquid side balance where we immediately realise
that the continuous cultivation will require all terms of the equation.

𝑑(𝑐 𝑦,𝑜𝑢𝑡 ∗ 𝑉)
= 𝐹𝑖𝑛 ∗ 𝑐 𝑦,𝑖𝑛 − 𝐹𝑜𝑢𝑡 ∗ 𝑐 𝑦,𝑜𝑢𝑡 ± 𝑟 𝑦 ∗ 𝑉 (8.14)
𝑑𝑡

The balance of cell mass in a chemostat

Let us first consider the mass balance for cells. Since there are no cells in the ingoing medium, and since the
volume is constant and 𝑟 𝑥 = 𝜇 ∗ 𝑐 𝑥 , the following equation is obtained:

𝑑𝑐 𝑥
𝑉∗ = −𝐹𝑜𝑢𝑡 ∗ 𝑐 𝑥 + 𝜇 ∗ 𝑐 𝑥 ∗ 𝑉 (8.15)
𝑑𝑡
or
𝑑𝑐 𝑥 𝐹
= − ∗ 𝑐 𝑥 + 𝜇 ∗ 𝑐 𝑥 (𝑑𝑦𝑛𝑎𝑚𝑖𝑐 𝑒 𝑥𝑝𝑟𝑒 𝑠𝑠𝑖𝑜𝑛) (8.16)
𝑑𝑡 𝑉

This is a dynamic expression which is valid at any time in the cultivation. This is in contrast to the steady
state expression which is derived below which is only valid when we have reached the desired operating
conditions. As we can see from the equation above, such a “steady state” situation is achieved only when
there is a perfect balance between the right-hand side components i.e. the flow (the first term) and the
growth (the second term) so that 𝑑𝑐 𝑥 /𝑑𝑡 = 0 and we will have a constant cell mass with time in the reactor.
In such a situation, the following expression is obtained for the specific growth rate at steady state in a
continuous cultivation operation (process operating point):

𝐹
𝜇= = 𝐷 𝑠𝑡𝑒 𝑎𝑑𝑦 𝑠𝑡 𝑎𝑡𝑒 (8.17)
𝑉

where D=F/V is called the “dilution rate”. The unit is naturally the same as for the specific growth rate: [h-1].
The specific growth rate in continuous cultivation at steady state is thus set only on the basis of the outgoing
flow rate at each operating point when a medium inflow has been chosen.
The inverted parameter V/F is the “residence time”, i.e. the average time that a given cell or molecule stay in
the reactor before being washed out.

𝑉
𝜏𝑅 = [𝑚𝑖𝑛] (8.18)
𝐹
110 8 Cultivation Concepts

This is an important concept since it indicates how long time must elapse after the conditions in the ingoing
flow are changed and until a new steady state is reached. If we wish to remove a component “A” in the
ingoing flow for example, four residence times is equivalent to a washout of 98.2% of the initial amount of
“A”. This is based on the time for exponential declination of the concentration of the medium in a perfectly
mixed reactor after a change has been made,

𝑟𝑒 𝑠𝑝𝑜𝑛𝑠𝑒 = 𝑒 −𝜙

where 𝜙 is a normalised residence time with respect to the concentration of the pulse/response. After four
mixing times this correspond to 1 ∗ 𝑒 −4 = 0.982 or equal to 98.2% washout of the pulse.

Mass balance of the limiting substrate

Let us next consider the mass balance on substrate. The substrate is present in all the terms of the general
equation and thus crosses the control surface in all the ingoing and outgoing flows. The substrate is consumed
(hence the minus sign), but the volume is constant and therefore:

𝑑𝑐 𝑠,𝑜𝑢𝑡
𝑉∗ = 𝐹𝑖𝑛 ∗ 𝑐 𝑠,𝑖𝑛 − 𝐹𝑜𝑢𝑡 ∗ 𝑐 𝑠,𝑜𝑢𝑡 − 𝑟 𝑠 ∗ 𝑉 (8.19)
𝑑𝑡

As before, 𝑟 𝑠 = 𝑞 𝑠 ∗ 𝑐 𝑥 , and after rearrangement the expression for the rate of change of the substrate
concentration under dynamic conditions is obtained:

𝑑𝑐 𝑠,𝑜𝑢𝑡 𝐹
= (𝑐 𝑠,𝑖𝑛 − 𝑐 𝑠,𝑜𝑢𝑡 ) − 𝑞 𝑠 ∗ 𝑐 𝑥 𝑑𝑦𝑛𝑎𝑚𝑖𝑐 (8.20)
𝑑𝑡 𝑉

At steady operating conditions, i.e. at steady state when 𝑑𝑐 𝑠 /𝑑𝑡 = 0, and the substrate concentration in the
bioreactor will not change, the following expression is obtained:

𝐹
(𝑐 𝑠,𝑖𝑛 − 𝑐 𝑠,𝑜𝑢𝑡 ) = 𝑞 𝑠 ∗ 𝑐 𝑥 𝑠𝑡𝑒 𝑎𝑑𝑦 𝑠𝑡𝑎𝑡𝑒 (8.21)
𝑉

This means that at steady state, the difference between the inflow and outflow of glucose (left-hand side)
balances the consumption (right-hand side). Note also that it is this particular steady state equation (substrate)
that is used for calculation of the ingoing flow to the reactor at a chosen substrate concentration in the ingoing
flow.

The condition for growth limitation

We noted earlier that we use continuous or fed-batch cultivation to reduce a specific substrate consumption
below the maximum to restrict the growth. The equation above enables us also to determine the general
condition which must be fulfilled to have a reduction in the growth rate, and we note that the flow of
substrate must be less than the value leading to the maximum consumption rate.

𝐹
(𝑐 𝑠,𝑖𝑛 − 𝑐 𝑠,𝑜𝑢𝑡 ) < 𝑞 𝑠,𝑚𝑎𝑥 ∗ 𝑐 𝑥,𝑚𝑎𝑥 (8.22)
𝑆

We defined the dilution rate before as F/V=D. If the qs -value is replaced by the value for the specific growth
rate by use of the definition of the biomass yield coefficient, i.e. 𝑌𝑥𝑠 = 𝜇/𝑞 𝑠 , we obtain:

𝜇
𝐷(𝑐 𝑠,𝑖𝑛 − 𝑐 𝑠,𝑜𝑢𝑡 ) = ∗ 𝑐𝑥 (8.23)
𝑌𝑥𝑠
8.3 Continuous cultivation 111

Since 𝜇=D, this gives the following expression for the cell concentration in the reactor at steady state:

𝑐 𝑥 = 𝑌𝑥𝑠 (𝑐 𝑠,𝑖𝑛 − 𝑐 𝑠,𝑜𝑢𝑡 ) 𝑠𝑡𝑒 𝑎𝑑𝑦 𝑠𝑡 𝑎𝑡𝑒 (8.24)

Since the yield coefficient is constant except at very low substrate concentrations (high concentrations are not
used), it is evident that the concentration of the cell mass at steady state depends only on the difference between
the inlet and outlet concentration of the limiting substrate at the operating point.
We have now derived the expressions for 𝜇 and 𝑐 𝑥 and to obtain an expression for the substrate concentration
in the bioreactor at steady state, we can use the Monod equation which is applicable at any substrate
concentration, also the very low, and noting that if 𝜇=D this gives:

𝑐𝑠
𝐷 = 𝜇𝑚𝑎𝑥 ∗ (8.25)
𝐾𝑠 + 𝑐𝑠

If this is solved for the substrate concentration in the reactor, 𝑐 𝑠 , we obtain the expression:

𝐷 ∗ 𝐾𝑠
𝑐𝑠 = 𝑠𝑡𝑒 𝑎𝑑𝑦 𝑠𝑡 𝑎𝑡𝑒 (8.26)
𝜇𝑚𝑎𝑥 − 𝐷

indicating that the substrate concentration at steady state is a function only of the dilution rate (the liquid
flow) and of the Ks -value of the substrate. This means that the higher the saturation constant, i.e. the lower
the affinity for the substrate, the higher is the apparent concentration in the reactor. Note that the substrate
concentration in the reactor does not depend on the substrate concentration in the inlet liquid flow which is
not a part of the steady state equation!

The mass balance of other substrates than the limiting

All substrates, except for the one chosen as limiting (equations above), are (must be!) present in non-limiting
amounts in the continuously operating reactor. Otherwise, the concept would not work. The mass balance for
a “substrate No 2” is the same as for all matter and thus analogous to that for the limiting substrate:

𝑑𝑐 𝑠 2,𝑜𝑢𝑡 𝐹
= (𝑐 𝑠 2,𝑖𝑛 − 𝑐 𝑠 2,𝑜𝑢𝑡 ) − 𝑞 𝑠 2 ∗ 𝑐 𝑥 (8.27)
𝑑𝑡 𝑉

Since the overall consumption rate is not maximal, the limitation of the growth also steers this consumption.
To calculate the concentration of the second substrate, an expression for the specific consumption rate is
required, and we can again use the yield concept for growth for this particular substrate:

𝜇
𝑞𝑠2 = (8.28)
𝑌𝑥𝑠 2

which can be inserted into the equation above giving:

𝑑𝑐 𝑠 2,𝑜𝑢𝑡 𝜇 ∗ 𝑐𝑥
= 𝐷(𝑐 𝑠 2,𝑖𝑛 − 𝑐 𝑠 2,𝑜𝑢𝑡 ) − (8.29)
𝑑𝑡 𝑌𝑥𝑠 2

The set-up of a continuous process in practice is shown in Figure 8.4.


112 8 Cultivation Concepts

Figure 8.4: Continuous cultivation in a 3L reactor with 1.5 L working volume. In this case, a 600L vessel is used as the container for the ingoing flow.
This is dimensioned to last for 6 weeks of cultivation for steady state growth of E.coli at a growth rate of 0,4 h-1. The volume is kept at steady state by
the pump for the outgoing flow which is here controlled by the weight of the reactor (note the reactor position on the balance). A foam layer is also seen
at the top of the liquid which is not desired but sometimes difficult to avoid

A plot of parameters against the dilution rate

In order to find a good operating point for chemostat cultivation, a plot of parameters against the dilution
rate is generally consulted. This is a curve in which a number of important parameters are plotted against the
dilution rate, D = F/V, for a specific medium and medium inflow concentration of the growth-limiting compound. A
plot of parameters against the dilution rate is shown in Figure 8.5:

Operating interval

cs,in
DOT
cx,cs [g/L]
DOT [%]

rx [gx/L,h]
Operating point
cx

rx
cs

D=F/V [h-1]

Figure 8.5: Continuous cultivation – a plot of parameters against the dilution rate .The vertical blue line indicates one of the possible steady state
operating points and the operating values are found where the line crosses the graphs of the different parameters. In the example chosen, the cell mass
(X) is 5 g/L, DOT is 80% and the volumetric reaction rate (𝑟 𝑥 ) is approximately 0.15 g/L,h at this point. The substrate concentration cannot be
estimated from the graph but will be in the mg/L region.

A plot of parameters against the dilution rate can be derived from practically performed cultivations run to
steady state at different dilution rates, D =F/V, with simultaneous measurement of the concentrations of
the interesting parameters. This curve can also be gained from a computer simulation provided the mass
balances and the in-going constants of these are known, or can be approximated, and that there are models
available for the kinetics of production/ consumption. Note again that the analysed parameters 𝑐 𝑥 , 𝑐 𝑠 , DOT
(or Dissolved Oxygen Tension which is a representative of the reactor concentration of oxygen in the medium,
more in the next chapter!) and 𝑟 𝑥 (calculated), are plotted in the plot of parameters against the dilution rate
not against time but as a function of the dilution rate that is the flow of liquid at each steady state. This means
that once the flow/volume is decided, all other parameters are fixed! This is what the plot tells and such
an operating point is shown by the vertical blue line in the graph. Each such vertical line represents a new
steady steady state which represents a different parameter set-up.
Let us consider the variables of the plot of parameters against the dilution rate as a function of D and compare
them with the equations just derived. We see that the cell mass concentration does not vary to any great
8.3 Continuous cultivation 113

extent with the dilution rate, which is good since we have postulated that it will depend only on the in-going
substrate concentration. Only at a very low dilution rate the flux to the maintenance requirement will become
dominant so that the cell mass concentration cannot be upheld (to the far left of the figure).
It is also evident that the substrate concentration in the reactor is a function of the dilution rate as we have
earlier shown by mass-balancing. Since the substrate concentration in the reactor changes with D, the specific
glucose consumption rate will also vary with D. If the consumption of sugar increases, so does the oxygen
consumption and this lowers the reactor oxygen “concentration” (DOT).
The volumetric production of cells is defined as 𝑟 𝑥 = 𝜇 ∗ 𝑐 𝑥 . Since it is known from the above that the cell
mass is constant but that the growth rate increases with D, 𝑟 𝑥 will thus also increase. If we desire a high
productivity, we need to use a point to the very right in the figure.

cs1
DOT2 DOT1
cx, cs [g/L]

cs2

DOT[%]
cx2
cx1

cs

D(h-1)

Figure 8.6: Comparison of two different inlet concentrations of glucose: S1 =6 and S2 =9 g/L and the effect on selected parameters. At the critical point,
all cells are washed out of the reactor and the substrate concentration will be equal to that of the incoming flow as usual in continuous cultivation

At a specific point at a very high dilution rate (indicated with the dotted orange line at the orange arrow),
corresponding to a high growth rate, the continuous cultivation becomes highly unstable and this cultivation
principle cannot be used. This occurs when we approach the maximum growth and substrate consumption
rates. At this point, all cells will be washed out of the reactor and the substrate concentration is equal to
that of the incoming flow since nothing is consumed (no cells). For this reason, the operating interval for a
continuous cultivation is limited to points well below the maximum growth rate. The reason, which we will
describe later on in detail, is that the internal control of the steady state, which is based on metabolic control
by the cell, cannot be upheld here.
The plot of parameters against the dilution rate above was based on a constant inlet concentration of glucose
i.e. for one specific medium composition. How would the curves appear if the glucose concentration in the
medium inlet were increased and we established a new steady state? From the mass balance on substrate, we
know that the cell concentration, 𝑐 𝑥 , will be increased by higher substrate concentration and consequently
also the volumetric growth rate, 𝑟 𝑥 . However, since the 𝑟 𝑥 is again directly proportional to the oxygen
consumption, 𝑟 𝑜 , an increase in cell mass requires more oxygen and the DOT (proportional to the oxygen
concentration in the medium) will also be affected. In other words, if the volumetric cell production rate, 𝑟 𝑥
[g/L,h], is increased, the oxygen consumption, 𝑟 𝑜 , will also increase and this will lower the DOT, which is
shown in Figure 8.6.
How will the other nutrients in the reactor vary with increasing inlet concentration of the limiting substrate?
An increase in e.g. the glucose concentration will lead to a higher cell concentration in the reactor, if this is
the limiting substrate, as we have determined before. More cells consume more substrate and this will lead to
lower values of all substrates. Care must thus be taken to ensure that the increased inflow of glucose does not
mean that another substrate becomes limiting, otherwise the control of the cultivation is lost. The effect on
other nutrients in the medium of increasing the inlet concentration is shown by Figure 8.7.
114 8 Cultivation Concepts

cs,in=5 g/L (inlet conc. of limiting cs,in=9.5 g/L (inlet conc. of


substrate) limiting substrate)

cx, cs, cs2 [g/L]


cx

cx
cs2 cs2
cs cs

D (h-1) D (h-1)

Figure 8.7: Effect of increased inlet substrate concentration on the consumption of other non-limiting substrates and the cell concentration

Reasons for restriction of the growth rate

What is the benefit of a continuous cultivation if it is not possible to radically increase the cell mass? Generally,
there are several reasons for restricting the growth rate and one is to avoid oxygen limitation. But a high
growth rate, meaning a high influx of glucose to the cell, also leads to overflow metabolism as shown earlier.
This has two major drawbacks: the first is that carbon goes to other products than new cells and the desired
product, which cannot be accepted in a yield sensitive process. The second drawback is that the accumulation
of overflow metabolites may reduce the growth rate since the overflow metabolite might be toxic to the cell.
This is true e.g. for E.coli cultivation and acetic acid production where the cells have to invest a lot of energy
to control the intracellular pH which is achieved by the pumping out of hydrogen ions.
The growth rate is also accompanied by the generation of heat since aerobic growth is an exothermal process.
If microorganisms grow too rapidly in too large numbers, it may be difficult to cool the reactors and maintain
the reactor temperature. The oxygen consumption, overflow and heat metabolism can mathematically be
linked to growth through the yield coefficients as shown in Table 8.1 for E. coli. Note particularly the close
coupling between the oxygen consumption and the heat generation. The figure “32” refers to the molar mass
of oxygen gas which is needed to compare, in this case, the molar quantities.

Table 8.1: Three important reasons for restriction of the growth rate. Note the coupling between this parameter and the acetic acid production, the heat
generation and the oxygen consumption

Effect to avoid Critical parameter Coupling of rates


Overflow metabolism Too high 𝑟 𝑥 = 𝜇 ∗ 𝑐 𝑥 𝜇 = 𝑞 𝑠 ∗ 𝑌𝑥𝑠 = 𝑞 𝐻𝐴𝑐 ∗ 𝑌𝑠𝐻𝐴𝑐 ∗ 𝑌𝑥𝑠
Heat generation Too high 𝑟 𝑥 = 𝜇 ∗ 𝑐 𝑥 𝜇 = 𝑞 𝑜 ∗ 𝑌𝑥𝑜 = 𝑞 𝐻 ∗ 1/𝑌𝐻𝑂 ∗ 32 ∗ 𝑌𝑥𝑜
Oxygen limitation Too high 𝑟 𝑥 = 𝜇 ∗ 𝑐 𝑥 𝜇 = 𝑞 𝑜 ∗ 𝑌𝑥𝑜

Taken together, this means that the chemostat can be used to prolong a cultivation in time since we can
overcome some of the batch concept limitations such as overflow metabolism, excessive heat generation and
anaerobic conditions due to too high an oxygen consumption.

The chemostat performance

Let us now consider the parameters in a chemostat as a function of time i.e. a principal plot of the cultivation
as it proceeds during production.
In the graph below, the chemostat is started from a batch cultivation. The left of the graph shows a typical
batch development pattern where the cell mass is increasing exponentially, in this case at the expense of the
8.3 Continuous cultivation 115

oxygen concentration level, DOT, which drops. The specific growth rate, 𝜇, is fixed at its maximum value, as
expected. To run this batch, a small amount of glucose was initially added to the medium but not so much
that it will exceed the desired cell concentration at steady state.
At a point when growth is steadily increasing, the feeding of the medium containing the growth-limiting
component is started. After the steady state is reached the volume is constant all the parameters will
remain constant with time. This is a state of metabolic balance when all compounds inside and outside
the cell will be constant with time. This means that the physiology remains the same at all times and this
includes also the product formation. And as we noted before, this cultivation can principally run forever, if
the medium reservoir is continually refilled. The values adopted at this point resemble those of the “blue
line” in the plot of parameters against the dilution rate.

[CONC] BATCH CHEMOSTAT

µ
DOT DOT ⋅cx⋅H
Y
DOT = DOT * − xo
kL a
cX cx=Yx/s*(cs,in-cs)

µ
Reduced F/V Increased si

µ
F
µ= =D
V
cX cS
D⋅ K s
cs =
µmax−D

TIME (h)

Figure 8.8: Chemostat operation with time. Changes are made in two different parameters as shown by arrows: left arrow: a reduction in the inlet flow
and right arrow: an increased concentration of the limiting glucose concentration in the inlet flow. To the right, the governing steady state expressions
are shown. The expression for the dissolved oxygen (DOT) is derived in chapter 9

To check the validity of our previously derived equations, let us decrease the inlet flow rate, i.e. the dilution
rate (left arrow in the graph). All the parameters that depend on this rate, i.e. 𝜇, 𝑐 𝑠 and DOT should change
their values. We see that this is indeed the case but in reality this may take four residence times or more since
it will take some time for the new situation to develop. Since the growth rate declines, so does the oxygen
consumption. Less oxygen is used, and the DOT value increases. However, this does not change the cell mass
concentration, as is also evident from the graph. The corresponding steady state balances are shown to the
right in the figure. The equation for consumption of dissolved oxygen will be developed in chapter 9.
The second check is to increase the substrate concentration of the growth-limiting substance; in this case
glucose (right arrow in the graph). This means that we change the medium and we cannot use the same plot
of parameters against the dilution rate since this is dependent on a fixed limiting substrate concentration.
According to the derived steady state equations this will lead to an increase in cell mass concentration and
DOT but will not affect the other parameters. Also here some parameters will temporary vary (in particular 𝜇
and 𝑐 𝑠 ) until new steady state conditions are established.

Effects of the saturation constant

The cell affinity for a particular substrate has a large effect on the developed concentration of this particular
nutrient in the bioreactor. We might chose e.g. glucose, glycerol or other sugars, which might have largely
different Ks values as we saw in a preceding chapter. We have already derived an expression for the reactor
116 8 Cultivation Concepts

concentration of sugar from the Monod equation as:

𝐷 ∗ 𝐾𝑠
𝑐𝑠 = (8.30)
𝜇𝑚𝑎𝑥 − 𝐷

This value we also refer to as the “residual substrate concentration” i.e. the value which represents the
substrate that leaves the reactor with time. We see from the equation that he higher the Ks -value the higher is
this concentration. A simulation of the plot of parameters against the dilution rate for substrates that have
10-times differences in saturation constant shows the following:

increased affinity

Ks=1 [g/L] Ks=0.1 [g/L] Ks=0.01 [g/L]


cS cS
cX, cs [g/L]

cS
cX cX cX

cS cS cS
D [h-1] D [h-1] D [h-1]

Figure 8.9: Effect of the saturation constant on the residual glucose concentration

This concentration is particularly important since the outflow should not contain too high an amount, since a
lot of nutrients are then lost and further constitutes a pollutant in the waste water. It is evident in the graph
that the higher the affinity (the lower the saturation constant), the lower is the residual concentration. This
means that a substrate like glucose is to be preferred at least from this respect.

The self-regulating metabolic control of the chemostat

The self-regulating metabolic control of the chemostat It is evident that there is a certain critical value of
the dilution rate above which the chemostat cannot run. All other operating points are however available.
Provided that the volume is constant, i.e. that the inlet and outlet liquid flows are balanced, there will be a
balanced growth and a steady state will develop. This means in principle that each new cell produced in
the bioreactor at a certain rate is washed-out by an equal flow-through of fresh substrate and this allows all
parameters to remain at the same value.
In practical applications, however, a perfect balance cannot always be achieved. Pump speeds are notoriously
known to fluctuate with time and there may also be small fluctuations in the limiting nutrient concentration.
Under such conditions a deviation from the steady state will follow. To understand what happens, the
dynamic balances, rather than the steady state solutions, must be considered. The dynamic mass balance on
cells provides the useful expression to start from.

𝑑𝑐 𝑥 𝐹 𝐹
 
= 𝜇 ∗ 𝑐𝑥 − ∗ 𝑐𝑥 = 𝑐𝑥 𝜇 − (8.31)
𝑑𝑡 𝑉 𝑉

This expression shows clearly that a balance is necessary between the liquid flow (F/V) and the speed at
which the cells actually grow (𝜇) when we have steady state.
Let us consider what happens if we leave a steady state where 𝐹/𝑉 = 𝜇 (𝑑𝑐 𝑥 /𝑑𝑡 = 0) and there is a temporary
increase in the dilution rate, i.e. the inlet flow, F, increases. This means that F/V > 𝜇 and the rate of
8.3 Continuous cultivation 117

accumulation of cells becomes negative, 𝑑𝑐 𝑥 /𝑑𝑡 < 0. The substrate consumption rate, 𝑑𝑐 𝑠 /𝑑𝑡 , on the other
hand, increases with increased F according to the mass balance equation on glucose and 𝑑𝑐 𝑠 /𝑑𝑡 > 0.

𝑑𝑐 𝑠,𝑜𝑢𝑡 𝐹
= (𝑐 𝑠,𝑖𝑛 − 𝑐 𝑠,𝑜𝑢𝑡 ) − 𝑞 𝑠 ∗ 𝑐 𝑥 (8.32)
𝑑𝑡 𝑉

This seems logical since a temporary lower cell mass accumulation rate will mean that there are fewer cells to
consume the sugar, so that a temporary substrate accumulation take place. According to the Monod equation,
this will however, in turn, lead to an increase in the consumption rate and more cells will be created and we
are thus back at steady state! The physical imbalance in the liquid inflow has thus resulted in a metabolic
effect that leads to an increase in the cell concentration, i.e. an act of metabolic balancing. The opposite occurs
if there is a temporary decrease in the dilution rate.
These two examples show the unique and cell-internal and self-controlling ability that characterises the
chemostat but no other cultivation concept.
This is all very well but there is a limit to where such disturbances can be metabolically controlled. The
disturbance in the inlet flow, i.e. the temporary increase in the flow, cannot be so great that it becomes
impossible for the cells to increase their growth rate which obviously has its maximum value. Since the cells
cannot grow more rapidly, they will be washed out (see the plot of parameters against the dilution rate).

Applications of the chemostat

Despite the fact that a very high cell mass cannot be accumulated in the chemostat, the concept still allows
the highest volumetric productivity, r [kg/L,h], of all cultivation concepts. This is because the process does
not have to be temporarily shut down as in batch and fedbatch cultivation. The effect is particularly clear at
high growth rates (to the right in Figure 8.10) and when long batch refilling times are needed.

Figure 8.10: Effect of batch reactor downtime on productivity in comparison with continuous cultivation. The downtime is the time needed between
batches to prepare the reactors, equipment and the propagation of a new inoculum which is not needed in continuous cultivation

The continuous cultivation concept is not very extensively used in the industry. Two important examples
of continuous industrial cultivation however are the two very different processes for biological wastewater
treatment and the production of recombinant insulin in yeast by Novo Nordisk. However, the concept can in
principle be used for most microbial processes. The chemostat system is however extensively used in research
because it allows constant physiological conditions at all times since all the parameters are constant i.e. in
steady state. This mode of operation can thus be used for determination constants in kinetic models and for
deriving specific yield coefficients.
118 8 Cultivation Concepts

The determination of the saturation constant of the Monod equation is one such example. To determine this, a
number of steady states (operating points) are set up in the chemostat and the limiting substrate concentration
is measured at each point. This results in a data set of specific growth rates and the corresponding substrate
concentrations which are related by the Monod equation:

𝑐𝑠
𝜇 = 𝜇𝑚𝑎𝑥 (8.33)
𝐾𝑠 + 𝑐𝑠

To derive the saturation constant the equation can be rewritten as a straight line, as shown in Figure 8.11.
After rearrangement it is seen that a plot of 1/𝜇 against 1/cs gives a slope and intercept with the x-axis that
can be used to calculate the coefficients in the equation.

x
x
x Ks
1 x max
- x
Ks
1
cs
Ks

Figure 8.11: Inverse Lineweaver Burke plot to determine the Monod equation coefficients. The red “x:es” show the derived steady-state values

The continuous process has however some drawbacks of different character:

Table 8.2: Drawbacks and comments on the continuous cultivation concept

Often cited drawbacks Comment


Operating skills needed A one-time effort
Not so drastic but different systems for volume control can
Laborious equipment
take time to establish
Not operable at maximum growth
Not really necessary for most processes
rate
Wall growth of cells Not all organisms, but difficult to overcome
Impossible to use unstable cells, but maybe these are anyway
Genetic instability
unwanted?
Increased risk of infection High risk of outgrowth by infection if something grows faster
High residual substrate concentra-
True in some cases, but importance depends on the process
tion
Limited to growth-related prod-
But many products are! (i.e. primary metabolites and cells)
uct formation

The genetic instability of the chemostat is probably a major reason why this cultivation technique is not often
used. Due to the long cultivation time many new generations of cells are produced. The cells may eventually
adapt to the prevailing conditions by mutation and metabolic parameters will change. One example is the
parameters of the Monod equation. Here the substrate uptake may with time become increasingly efficient so
that both the affinity and the maximum rate of the glucose uptake i.e. the capacity of the cell wall permeases
may change. This will give an unstable process and if this negatively interferes with the product formation
the process will not be scaled up to production.
8.4 Fedbatch cultivation 119

The risk of infection is also serious, and measures must be adopted to ensure that no other organisms slip
into the reactor. Usually, a slight over-pressure in the reactor is used for this purpose and extreme care is
taken when medium reservoirs are changed. All reactor ports are also frequently controlled and cleaned, and
in-going and outgoing filters are sometimes also continuously sterilised by steam.

Mutant cX2
cX2
cX1
cX
cS

Figure 8.12: Up-growth of a single cell mutation with higher specific substrate consumption rate than the production cell at time=0 and the
simultaneous washout of the production cell

Figure 8.12 shows the effect of introducing a single mutant cell to the reactor at time = 0 in a chemostat and
the outgrowth of the original cell after 250 h. The only difference between the cells is the maximum specific
glucose uptake rate, qs , which are 1.5 and 2.0 g/g,h, respectively.

Continuous cultivation with recirculation

The idea of continuous cultivation with recirculation was introduced as an easy way to increase the cell
concentration in the reactor while still being able to operate at steady state and thus to have a continuous
production process. A common application is animal cell cultivation where the concept is called “perfusion
culture”. The cells are retained or recirculated back to the bioreactor by separation with e.g. a filtration
device.
Recirculation is also used in the activated sludge process in wastewater treatment where excess sludge is
pumped back to the reactor from a sedimentation tank to increase the rate of microbial conversion of carbon
and nitrogen substances. The use of immobilized cells growing on beads in a bioreactor is a third example,
which mathematically resemble a continuous cultivation with recirculation.

8.4 Fedbatch cultivation

Both the continuous and the fed-batch cultivation concepts were first presented at the end of the 19th century
and the patents for this technique were granted at the beginning of the 20th century. The reason for the
fedbatch development was that some commonly used products could not be produced in amounts sufficient
to satisfy the customer demand. One of the earliest fed-batch process examples is the production of baker’s
yeast. Here, the cell itself is the product and a high cell mass per unit time is consequently needed.
The continuous cultivation concept provides many advantages over the batch concept, but with neither of
these two concepts is it possible to increase the cell mass to a very high density. Fed-batch is, as we shall see,
the only cultivation concept that makes this possible.
120 8 Cultivation Concepts

It has already been stated that in order to reduce the growth rate it is necessary to restrict the substrate feed
of a limiting component, usually the carbon source and often glucose. The general limitation criterion was
earlier given:

𝐹𝑖𝑛 ∗ 𝑐 𝑠,𝑖𝑛 < 𝑞 𝑠,𝑚𝑎𝑥 ∗ 𝑐 𝑥 (𝑡) ∗ 𝑉(𝑡) (8.34)

The development of the cell mass is obviously time-dependent and this is also true of the liquid volume
which increases in fedbatch as we shall see. How can we operate the reactor so that we can now increase the
cell mass with time? The answer is that this requires a concomitant reduction of the substrate flow to the
reactor with time and it cannot be constant as in the case of continuous cultivation. The following figure is
drawn to illustrate this.

Figure 8.13: Cell concentration as a function of time at different growth rates which are shown by five different curves. Two points are highlighted
where the volumetric cell accumulation rates (slope of the curve), 𝑟 𝑥 = 𝑑𝑐 𝑥 /𝑑𝑡 , or 𝜇 ∗ 𝑐 𝑥 , are the same. If a higher cell mass is desired, a lower growth
rate must follow!

The figure shows the cell growth with time from cultivation at four different growth rates. Two points
are highlighted at the curves of the highest and the lowest growth rates, respectively. At these points, the
volumetric accumulation of cells with time ( 𝑑𝑐 𝑥 /𝑑𝑡 ) is the same (the slope) but we see also that for the high
growth rate, this takes place at a considerably lower cell mass. These two points thus show the two possible
strategies for achieving the same productivity: by choosing either a low cell density and a high specific
growth rate or a high cell density and a low specific growth rate. This is because the oxygen consumption is
proportional to 𝑑𝑐 𝑥 /𝑑𝑡 = 𝜇 ∗ 𝑐 𝑥 , the volumetric rate, as is also the production of heat as was shown when we
discussed continuous cultivation. And since we now desire to have a high cell density, x, the only solution
is to restrict the specific growth rate, 𝜇, in such a way that it is continuously decreasing as the cell mass
increases.
In practice this is done in the same way as for the continuous cultivation i.e. by choosing a limiting substrate,
e.g. glucose, which is excluded from the reactor medium and is added to the cultivation in a manner which
ensures that the growth is at all times less than the maximum. However, since we do not wish to loose
any cells, there is no liquid outflow and everything stays in the reactor. This means that the liquid volume
increases and this will eventually limit this particular process. In order to hold back the increase in volume,
we feed just the limiting component and not the complete medium. This is fed in very high concentrations,
usually as a 60% solution, i.e. close to 600 g/L, to avoid a high volume build-up.
The stages to implement the fedbatch process are basically the same as in batch cultivation except for the feed
to the reactor over time (Figure 8.14).
It is easy to realise why the word “batch” is retained also for this concept since it is in practice as work-intensive
as the batch concept. Since the volume increases, only a limited time can elapse before it is necessary to stop
the cultivation and start all over again.
8.4 Fedbatch cultivation 121

Figure 8.14: Cell concentration as a function of time at different growth rates which are shown by five different curves. Two points are highlighted
where the volumetric cell accumulation rates (slope of the curve), 𝑟 𝑥 = 𝑑𝑐 𝑥 /𝑑𝑡 , or 𝜇 ∗ 𝑐 𝑥 , are the same. If a higher cell mass is desired, a lower growth
rate must follow!

Mass balances in fedbatch cultivation

The fed-batch system can be described in the same way as the other concepts, i.e. by mass balancing over a
control volume.

Figure 8.15: Control volume used for fedbatch cultivation with no outflow

Here we see that the situation is apparently simpler than for continuous cultivation since there is only an
inflow term and no outflow. We say “apparent” since we note that the volume changes with time in this case.
The general balance for any component “y” in the fed-batch system is thus:

𝑑(𝑐 𝑦 ∗ 𝑉)
= 𝐹 ∗ 𝑐 𝑦 ± 𝑟𝑦 ∗ 𝑉 (8.35)
𝑑𝑡

The balance for cell mass and product in fedbatch cultivation

The mass balance for cells is clearly very similar to the one for batch cultivation since there is no inflow of
cells. The same is true for the product. Please note however that since the volume is not constant this radically
changes the outcome.

𝑑(𝑐 𝑥 ∗ 𝑉)
= 𝑟𝑥 ∗ 𝑉 (8.36)
𝑑𝑡

𝑑(𝑐 𝑝 ∗ 𝑉)
= 𝑟𝑝 ∗ 𝑉 (8.37)
𝑑𝑡
122 8 Cultivation Concepts

Developing this expression using partial differentiation for cell accumulation leads to:

𝑑𝑉 𝑑𝑐 𝑥
𝑐𝑥 +𝑉 = 𝑟𝑥 ∗ 𝑉 (8.38)
𝑑𝑡 𝑑𝑡

and since 𝑑𝑉/𝑑𝑡 = 𝐹𝑖𝑛 − 𝐹𝑜𝑢𝑡 but 𝐹𝑜𝑢𝑡 = 0, rearrangement yields:

𝑑𝑐 𝑥 𝐹
= − ∗ 𝑐 𝑥 + 𝑟𝑥 (8.39)
𝑑𝑡 𝑉

and we see that the accumulation of cells will at all times be diluted by the inlet flow. A similar operation
leads to the expression for the product formation:

𝑑𝑐 𝑝 𝐹
=− ∗ 𝑐 𝑝 + 𝑟𝑝 (8.40)
𝑑𝑡 𝑉

The mass balance for substrate in fedbatch cultivation

This mass balance is slightly more complicated since there is also an inflow of substrate (and a negative
consumption term as usual):

𝑑(𝑐 𝑠 ∗ 𝑉)
= 𝐹 ∗ 𝑐 𝑠,𝑖𝑛 − 𝑟 𝑠 ∗ 𝑉 (8.41)
𝑑𝑡

The differential is again developed in the same manner as above and we get:

𝑑𝑐 𝑠 𝐹
= (𝑐 𝑠,𝑖𝑛 − 𝑐 𝑠 ) − 𝑟 𝑠 (8.42)
𝑑𝑡 𝑉

We note that all three expressions are identical to those used for continuous cultivation! This is due to that
the dilution due to the flow-through of fresh medium in a continuously operated reactor, is in fedbatch
replaced by an internal dilution due to the in-going flow. Although these processes are radically different,
mathematically there is however no difference.
For the fedbatch equations there is no analytical solution available and a numerical method must be applied.
To do so, we must introduce some expressions for the volumetric formation/consumption rates (qs and qp )
and the kinetic expressions that we have seen in former chapters, (e.g. Monod, Luedeking-Piret etc.) can
become very useful provided they can describe the situation in a particular process.

The substrate feed

The design of fedbatch cultivation relies first and foremost on the setting of a so-called “feed profile”. The
trick with the fed-batch cultivation, which allows the accumulation to high cell density, lies in the design of a
profile whereby the feed of substrate to each individual cell is continuously reduced. How is this done?
In industrial cultivation, several different feed profiles can be used, but the traditional processes are all based
on the same theme and the principal outline is that:

I The process is started with an exponential feed of the limiting substrate up to the point of reactor
capacity for oxygenation and heat removal and the feed is thereafter set at a constant value
8.4 Fedbatch cultivation 123

Let us explore how this feed profile setting works into some detail.
Phase 1: An exponential feed will allow the cells to grow exponentially but we will not set it to the maximum
growth rate, since even a slight overfeed lead to substrate accumulation and the development of batch
conditions. A too high feed will also lead to accumulation of overflow metabolites which are in most cases
undesirable. The following simple equation for the exponential flow is used where 𝜇 < 𝜇𝑚𝑎𝑥 :

𝑑𝐹
=𝜇∗𝐹 (8.43)
𝑑𝑡

The specific growth rate can be set to any value but is generally chosen as high as possible but low enough
to a point where earlier knowledge indicates that no overflow metabolites are formed. The exponential
feed results in a constant specific growth rate during this feed phase. This makes it possible to integrate
the equation from t = 0 to t and from F0 at the start to F, and thus to obtain the following expression for a
time-dependent exponential flow rate:

𝐹(𝑡) = 𝐹0 ∗ 𝑒 𝜇∗𝑡 (8.44)

In practice, the software for process control many times allows the direct introduction of this expression
for feed control and a smooth exponential increase in the flow rate develops over time. For simpler control
systems however, step-wise changes in the feed can only be allowed and it is necessary to calculate these
individual steps where the flow rate is constant for a certain time and then increases. The software generally
allows about 10-20 discrete steps in the inflow over the whole process. The amount of liquid added with both
systems will of course eventually be the same but, with the second approach, the growth curve is not as
smooth as with the more sophisticated control algorithms and there is a risk that cells may react to abrupt
changes in the substrate availability.
During this initial exponential feed phase, the growth rate is constant while the cell mass increases
exponentially. This means that the value of the volumetric productivity, 𝑟 𝑥 = 𝜇 ∗ 𝑐 𝑥 , increases and, since the
growth rate is proportional to the oxygen consumption, this feed profile can only run for a limited time, until
a point where the reactor oxygen transfer capacity is reached, or in practice, slightly earlier than this to be on
the safe side. This is the critical value that constitutes the switch point for the feed in fedbatch cultivation.
As we said above, the exponential feed results in a constant growth rate (although not the maximum). Note,
that this is actually a situation which is characterised by a steady state in all variables except for the cell mass
and volume but it will only last for a limited time. This is however often used in research to replace the
steady-state operation in continuous cultivation which can be both technically and practically more difficult
to design.
Phase 2: After the critical point where the reactor oxygen transfer capacity has been reached (or heat removal
capacity, of course!), it is in principle very easy to steer the flow so that the cell mass continues to increase but
without the critical limitations. This is again realised from the volumetric growth rate expression, 𝑟 𝑥 = 𝜇 ∗ 𝑐 𝑥 .
If the cell mass, 𝑐 𝑥 , is to increase, the growth rate must decrease to keep 𝑟 𝑥 constant, and this can easily be
achieved by applying of a constant feed rate. If the feed does not change and the cells continue to grow in
number, each cell will get a progressively smaller amount of substrate with time. According to the Monod
equation, this will lead to a lower growth rate. So, the high growth rate is sacrificed for a high cell density.
In practice, this is done by changing the controller from step-wise increased additions to a stop at the last
derived value which is kept over the last part of the cultivation.

𝐹(𝑡) = 𝑐𝑜𝑛𝑠𝑡 𝑎𝑛𝑡

In summary, the design of the common industrial feed profile is thus generally a two-phase operation with
an exponential feed phase followed by a constant one.
124 8 Cultivation Concepts

Another start-up of the fedbatch is possible when the accumulation of overflow metabolites is unimportant
or very low. In this case, the fedbatch may start from a batch cultivation as seen below.

batch phase constant feed phase


1 20 4 0.2

cs
cx
(µ g/g,h)

cx (g/L)

cs (g/L)

F (L/h)
F

µ
0 0 0 0
0 2 4 6 8 10 12
time (h)
time (h)
Figure 8.16: Start of a fedbatch cultivation from phase 1: a batch cultivation with 10 g/L glucose which is followed by phase 2: a constant feed phase.
[s]=concentration of substrate, [x]= cell mass concentration and 𝜇=the specific growth rate. The substrate concentration (in red) is partly plotted
outside the figure to be able to show the low concentrations during the constant feed phase

The batch protocol is designed for the addition of a restricted amount of the limiting substrate which is
chosen to give a cell mass where the volumetric production rate of cells, rx, (the product of cells times the
specific growth rate) does not exceed the reactor capacity for oxygen transfer. This is principally the same as
for setting of an exponential feed although here we operate at the maximum growth rate. However, when we
use the batch as a starting phase, the final cell concentration will consequently be lower than if the fedbatch
is started with an exponential feed. However, the critical point is reached earlier in the process which saves
expensive reactor time and generally increases the productivity.
Regardless of the choice of a batch or an exponential feed start in the first phase, the constant feed phase
develops in a similar way. The continuing increase of the cell mass leads to a decrease in the glucose
concentration which, in turn, leads to a decrease in the specific glucose consumption rate and consequently
also in a lower specific growth rate. The product of the cell mass concentration and the specific growth
rate, 𝑟 𝑥 = 𝜇 ∗ 𝑐 𝑥 , is during this phase constant, as are also the oxygen consumption and the microbial heat
production as we planned.
A positive issue with the batch feed start is that the operator easily detects the point when the glucose is
finished from the batch, since it results in a sharp increase in the DOT-value (no sugar-no respiration!) and it
becomes necessary to start a constant feed, which is illustrated in the simulation below.
The calculation of the constant feed rate is easily done from a mass balance on substrate and by introducing
𝑟𝑠 = 𝑞 𝑠 ∗ 𝑐 𝑥 :

𝑑𝑐 𝑠 𝐹
= (𝑐 𝑠,𝑖𝑛 − 𝑐 𝑠 ) − 𝑞 𝑠 ∗ 𝑐 𝑥 (8.45)
𝑑𝑡 𝑉

At the point when the glucose is finished, 𝑑𝑐 𝑠 /𝑑𝑡 = 0 as shown by the DOT signal. Since we seldom know
the specific rate of consumption of sugar, we go over to the expression for the specific growth rate through
8.4 Fedbatch cultivation 125

Figure 8.17: Start of a fedbatch cultivation from phase 1: a batch cultivation with 10 g/L glucose which is followed by phase 2: a constant feed phase.
[s]=concentration of substrate, [x]= cell mass concentration and 𝜇=the specific growth rate. The substrate concentration (in red) is partly plotted
outside the figure to be able to show the low concentrations during the constant feed phase

the expression for the yield of cells per glucose consumed i.e. 𝑌𝑥𝑠 = 𝜇/𝑞 𝑠 . The introduction of these terms
gives:

𝐹0 𝜇
(𝑐 𝑠,𝑖𝑛 − 𝑐 𝑠 ) = ∗ 𝑐𝑥 (8.46)
𝑉 𝑌𝑥𝑠
and
𝜇 ∗ 𝑐𝑥 ∗ 𝑉
𝐹0 = (8.47)
𝑌𝑥𝑠 ∗ 𝑐 𝑠,𝑖𝑛

The term “cs ” i.e. the substrate concentration in the bioreactor, is omitted since the concentration in the inlet
flow is very much higher than the concentration in the reactor, i.e. 𝑐 𝑠,𝑖𝑛 >> 𝑐 𝑠 , in fedbatch cultivation. In this
equation, all the values are known or quite easily approximated and they remain constant over a considerable
time.

Figure 8.18: Effect of the feed rate chosen for the exponential phase. Left: 𝐹 = 𝐹0 𝑒 0.3𝑡 . Right: 𝐹 = 𝐹0 𝑒 0.2𝑡 . More ethanol (in green) is formed at the
higher feed rate (left graph)

The exception is when we come to very low uptake values which may only be enough for maintenance
requirements. The yield coefficient, Yx/s , is often approximated to 0.5, since it is known that half the glucose
will go to new cell mass and half to energy formation and this is used since the actual value is often not
known. This is not however always true (as we have discussed before) and the maintenance consumption
reminds us of this at the end of the process. It is however often a good approximation.
The left-hand graph shows that the time taken to reach the constant phase is two hours shorter with the
higher feed. However, this also leads to a pronounced formation of ethanol, and the higher feed rate also
126 8 Cultivation Concepts

leads to a greater consumption of oxygen (DOT) which becomes dangerously close to zero. If this feed was
chosen, the exponential phase should probably, as a precaution, stop somewhat earlier.

batch phase constant feed phase


100 30 1

DOT (% air sat.)


cx
DOT
cx (g/L)

cs (g/L)
cs
0 0 0
0 2 4 6 8 10 12
time (h)

increase in feed

Figure 8.19: Effect on glucose concentration, DOT and cell accumulation of an increase in the feed to the reactor in a fedbatch cultivation. The
fedbatch is here started as a batch process (far left in the graph)

The feed rate of the limiting substrate (here glucose) influences its concentration in the reactor since there
is no outflow. The effect of a sudden increase in the feed on glucose concentration and on other important
parameters due e.g. to process instability, is shown in Figure 8.20.
The increase in the feed (red arrow) results in a higher glucose concentration, increases the substrate uptake
(Monod!) which increases the growth rate so that more cells are produced with time (in blue). The higher cell
concentration results in higher oxygen consumption and the DOT value is reduced as a consequence.

batch phase constant feed phase


100 30 1

DOT
DOT (% air sat.)

cx
cx (g/L)

cs (g/L)

cs
0 0 0
0 2 4 6 8 10 12
time (h)
Time [h]

Figure 8.20: Effect of the saturation constant on the glucose concentration in the fed-batch bioreactor. The arrow shows the decreasing ks -values used
in the simulation. From right to left the values are: 0.180; 0.100; 0.050; 0.010 g/L, respectively. A high Ks gives a lower specific consumption rate
according to Monod kinetics and this increases the DOT value since the need for oxygen is diminished

The actual value of the glucose concentration in the fedbatch reactor is determined from a numerical solution
of the mass balance on glucose. From a simulation we will see the dependence on the Ks -value. The higher
the saturation constant, the higher is the residual glucose concentration as was also the case in continuous
cultivation.
A fedbatch cultivation will thus be sensitive to fluctuations in the inlet flow and if the feed was to be
completely stopped, the glucose is rapidly consumed. The ks-value and the biomass concentration at this
8.4 Fedbatch cultivation 127

point affect the time before complete exhaust of the limiting component. We can see in Figure 8.21 that under
industrial conditions and at high cell densities (to the right-hand side of the figure) the available glucose
would be consumed in a couple of seconds in an E. coli cultivation at high cell density.

Figure 8.21: Effect of biomass concentration and saturation constant on the time taken for glucose exhaustion if the liquid inflow of glucose is stopped
for cultivation of two common microorganisms

The maximum cell density

The maximum cell concentration that can be achieved in a fedbatch cultivation is of course an interesting
design parameter. The glucose concentration decreases with time and the maintenance requirements will
therefore eventually become increasingly dominant and will eventually set a stop to further cell accumulation.
This can easily be seen from the mass balance on glucose:

𝑑𝑐 𝑠 𝐹
= (𝑐 𝑠,𝑖𝑛 − 𝑐 𝑠,𝑜𝑢𝑡 ) − 𝑞 𝑠 ∗ 𝑐 𝑥 (8.48)
𝑑𝑡 𝑉

To introduce the maintenance requirement, the previously derived expression for this concept is introduced:

𝜇
𝑞𝑠 = + 𝑞𝑚 (8.49)
𝑌𝑒𝑚

The maximum cell density is reached when the external glucose is exhausted, 𝑐 𝑠 = 0. This means of course
also that 𝑑𝑐 𝑠 /𝑑𝑡 = 0 and hence there is no growth i.e. 𝜇 = 0. Adding this to the mass balance leads to the
following expression:

𝐹
∗ 𝑐 𝑠,𝑖𝑛 = 𝑞 𝑚 ∗ 𝑐 𝑥,𝑚𝑎𝑥 (8.50)
𝑉

Solving this expression for the cell mass concentration leads to the following value for the maximum
concentration:

𝐹 1
𝑐 𝑥,𝑚𝑎𝑥 = ∗ 𝑐 𝑠,𝑖𝑛 ∗ (8.51)
𝑉 𝑞𝑚

The maximum cell density that can be achieved is thus dependent only on the inlet flow of glucose and
on the maintenance requirement, and it can be seen that a high maintenance requirement will, as may be
expected, reduce the maximum cell concentration.
128 8 Cultivation Concepts

Figure 8.22: Left: Effect of maintenance requirement on the maximum cell density. The maintenance value increases from top to bottom. Right: the
influence of the glucose inflow rate on the maximum cell density in a fedbatch cultivation. Top value: 14 g/h and bottom: 7 g/h
Aeration 9
Aeration is the process by which air is introduced into the bioreactor and 9.1 Parameters controlling the oxy-
the process by which oxygen is transferred from the gas phase (air) and genation . . . . . . . . . . . . . . . 129
dispersed in the liquid phase i.e. the medium. Many different parameters 9.2 The macroscopic mass balance of
oxygen . . . . . . . . . . . . . . . . 133
affect the efficiency of this process which is particularly critical for the
9.3 The molecular oxygen transfer
cultivation of many important aerobic microorganisms.
rate, OTR . . . . . . . . . . . . . . 134
In the following we shall for simplicity discuss only the transport of 9.4 Spargers . . . . . . . . . . . . . 141
oxygen but we realise at the same time that the removal of carbon dioxide 9.5 Methods to estimate the kL a 142
resulting from the respiration is equally important. The partial pressure The gas balance method . . . 142
Gassing out with nitrogen . 144
of carbon dioxide leads to growth cessation if it becomes too high and is
The dynamic method . . . . 145
actually commercially used as a means to hold back growth when it is
9.6 Mixing and aeration . . . . . 147
not desired e.g. in certain food preparations. If we want to reduce the
amount of carbon dioxide during cultivation, we increase the airflow
through the reactor. For some processes we need however to be careful
about ventilation of the carbon dioxide and that is immediately after the
inoculation. The reason is that the time taken for some cells to induce
anaplerotic carbon dioxide fixating reactions under certain medium
conditions might increase the lag phase that follows inoculation, if too
low concentrations are available. This is particularly important to realise
if we inoculate with a very low cell mass, since it will take time to build-up
the carbon dioxide concentration. Except for these particular effects of
this gas, the transfer properties of this exhaust gas are dependent on the
same mechanisms as the supply of oxygen and it is thus not separately
treated. A good aeration thus generally leads to a good carbon dioxide
removal.
In conclusion we may say that the concentration of both oxygen and
carbon dioxide are important and we may realise that these will vary
during the time course of cultivation. The few cells at the start of cul-
tivation will consume very little oxygen and produce low amounts of
carbon dioxide but when the cell mass increases the consumption of
oxygen becomes very high. This may lead to low oxygen concentrations
in particular at local points in the reactor which should be avoided, but
is in reality difficult to achieve. Research findings claim further that a too
high oxygen concentration should also not be avoided since it may hinder
growth. This is most likely to happen if pure oxygen is used probably
due to the formation of harmful oxygen radicals.

9.1 Parameters controlling the oxygenation

Oxygen is introduced as compressed air through a sparger in the biore-


actor and is thus introduced with a certain momentum. For a STR, the
sparger is positioned just below the lowest impeller, as shown in the ear-
lier chapter on bioreactors and for bubble columns these are positioned in
the reactor bottom. After leaving the sparger, bubbles of various sizes
130 9 Aeration

ascend in the reactor at a specific rate due to their low density and if the reactor is tall enough, a maximum
rising velocity (terminal velocity) is reached in the particular medium. The terminal velocity can be calculated
from a balance of the buoyant (positive) and the drag force (negative) where the net force is zero which
results in an expression called “Stoke´s law” where the velocity is depending on the following parameters:

𝑑𝑏2 ∗ 𝑔(𝜌 𝑙 − 𝜌 𝑔 )
𝑣∞ = (9.1)
18𝜂

We use this particular expression to realise that the terminal velocity is directly proportional to the density
difference between gas and liquid (𝜌) and the bubble size (db ) but inversely proportional to the liquid viscosity.
Different expressions of this law are used in various types of unit operations (e.g. in centrifugation).
We use a number of different parameters to describe the efficiency of oxygenation whereof some are derived
from the gas flow rate, Q [m3 /h], introduced to the reactor. In cultivation, the concept of VVM (volume
gas/volume liquid, minute) is very often used when we want to describe the degree of oxygenation per
volume liquid (Q/V). In a stirred tank reactor, this value is rarely greater than a factor of 1. To show the gas
flow in relation to the specific reactor dimensions, which is particularly important in bubble columns, the
inlet gas flow (m3 /s) is often divided by the cross-sectional area of the reactor to give the “superficial gas
velocity”, which is the linear gas rate over the reactor, often denoted by 𝑣 𝑔 (m/s).
Several of the factors that influence the transfer of gas from the bubbles are related to the medium. All
media have a certain ability to hold a gas and will also affect the form and size of the bubbles. The term “gas
hold-up” is defined as the volume fraction of gas that can be trapped by the liquid and which we want to be
high. The gas hold-up is expressed either in terms of the volume of the gas (VB ) plus liquid or of just the
volume of liquid (VL ). The first description is the more common and this leads to the following expression:

𝑉𝐵
𝜀= (9.2)
𝑉𝐵 + 𝑉𝐿

The presence of gas in the liquid will lead to a certain lowered density of this dispersion compared to pure
water and we realise that this may reach a local value of the gas hold-up in large reactors due to a fluctuating
flow. A bioreactor can contain a surprisingly large amount of gas and in a bubble column the value can be
close to 50%. The surprisingly high entrapment of gas in a liquid is illustrated in the Figure 9.1. Since the
introduction of gas into the liquid raises the surface level, this needs of course to be taken into account before
deciding the liquid volume.

Figure 9.1: Gas holdup in a bubble column. Note the different shapes of the bubbles!
9.1 Parameters controlling the oxygenation 131

To understand which parameters that may influence the gas hold-up we may use the following approximation.
The volume of gas in the reactor (VB ) we approximate by Q*𝜃 𝑔 where 𝜃 𝑔 is the mean residence time of the
gas in the reactor (or V/Q). The total volume (VB +VL ) is equal to A*H (area*height). We use these expressions
in the expression for gas hold-up and we get:

𝑉𝐵 𝑄 ∗ 𝜃𝑔 𝑄 ∗ 𝜃𝑔 𝑣𝑔
𝜀= ≈ = = (9.3)
𝑉𝐵 + 𝑉𝐿 𝑉𝐵 + 𝑉𝐿 𝐴∗𝐻 𝑣®

Finally, we identify the superficial gas velocity 𝑣 𝑔 = 𝑄/𝐴 and the mean rising velocity of a bubble as
𝑣 = 𝐻/𝜃 𝑔 from the equation. The gas hold-up is thus positively controlled by a high gas flow and negatively
controlled by a high mean bubble rise velocity which we get if we have large bubbles (Stoke´s law). However,
this has its limitations since at a very high gas hold-up, the bubbles may experience mutual contact and form
larger bubbles, which we want to avoid for several reasons as we shall see in the following.
The relationship between important parameters is often empirically determined to show effects on oxygenation.
The gas hold-up is often expressed as a function of the superficial gas velocity, as indicated above, for different
types of reactors. The bubble column has the highest gas hold-up followed by the stirred tank as shown
below. For this particular case, we see that the gas hold-up in the airlift reactor is only 1/3 of that in the
bubble column, since the aerated draught tube in this example occupies only 2/3 of the reactor in this
configuration.

Figure 9.2: Approximate gas hold-up in different reactor types. 𝑝 0 /𝑝 is a factor correcting for different pressures inside and outside of the reactor.

The medium used in the bioreactor also influences the mass transfer of oxygen since it also affects the size
of the bubbles as we have indicated above. In order to achieve a good mass transfer from a gas bubble,
the transfer area of the bubble (surface area) must be large in relation to its volume and consequently we
need small bubbles. This is shown in Figure 9.3, where we see that the larger the radius the smaller is the
surface/volume ratio.

Figure 9.3: Relation between surface area and volume


132 9 Aeration

How does the medium in fact affect the bubble size? There are several properties of a medium which will
influence the size of bubbles among which the surface tension, the liquid density and the character of the
medium components are the most important. The surface tension ( 𝛾 ) is created by cohesive forces among
the liquid molecules where each molecule is pulled equally in every direction making the net force=0. The
bubble size is directly proportional to the surface tension according to:

s
𝛾 ∗ 𝑑0
𝑑𝑏 = 1.7 ∗ 3
(9.4)
(𝜌 𝑙 − 𝜌 𝑔 )𝑔

where do is the size of the sparger orifice. The equation shows also that the design of the sparger is a most
important factor where the holes and how they are made control the initial bubble volume. The effect of the
density difference between the gas and medium is also evident. The surface tension decreases with lower
amounts of components in the medium to the low value of pure water. In water, bubbles are highly unstable.
The use of surfactants decreases further the surface tension which may lead to a stabilising effect of large
bubbles.
These properties will descide if a medium is “coalescent” or not i.e. if the gas bubbles are prone to merge
into larger bubbles. Pure water is a “coalescent” medium meaning that it tends to force bubbles to coalesce
into larger ones and this will reduce the surface/volume area. A cultivation medium containing salts, on the
other hand, is much less coalescent, leading to a higher mass transfer.
Another important feature of the medium is the solubility of oxygen (and carbon dioxide). The temperature
influences the solubility and it decreases with increasing temperature. A problem is that very little oxygen is
dissolved in water, only about 7 mg/L compared to the concentration in the gas which is approximately
300 mg/L. Air must therefore be continuously, and rapidly, supplied. The following example illustrates this
situation.

Example 9.1.1 We are cultivating cells when there is a sudden stop in the oxygen supply. We have 10 g/L
cells growing at a specific growth rate of 0.5 h-1 . The yield of cells per oxygen is 1 g cells per g oxygen. How
long will it take before the oxygen in the medium is finished?
The volumetric oxygen consumption rate is given by:

𝜇
𝑟 𝑜 = 𝑞 𝑜 ∗ 𝑐 𝑥 𝑎𝑛𝑑 𝑌𝑥/𝑜 =
𝑞𝑜

and this leads to:

𝜇 0.5 𝑔
 
𝑟𝑜 = ∗ 𝑐𝑥 = ∗ 10 = 5 (9.5)
𝑌𝑥/𝑜 1 𝑙, ℎ
This is the amount of oxygen that this amount of cells will consume in one hour. Since the concentration in
the medium is 7 mg/L, the oxygen is sufficient only for 5 seconds!

We may now summarise the most important factors that influence the oxygen transfer in a bioreactor as:

I The reactor geometry


I The gas flow
I The gas hold-up
I The bubble size (sparger configuration, medium properties)
I The mixing properties and presence of heterogeneity

Table 9.1 summary of the parameters that are generally connected to mass transfer of oxygen. Please observe
the units!
9.2 The macroscopic mass balance of oxygen 133

Table 9.1: Parameters connected to mass transfer of oxygen

Parameter Name Unit Abbreviation Comment


Gas velocity m3 Q
VVM Volume gas/volume VVM Obs! The rate is calculated
liquid, min in minutes! Very
commonly used!
Superficial gas velocity m/s Vg Linear gas rate Gas
hold-up
volume gas/total volume 𝜀 Ability of a liquid to hold
gas
Diffusivity m2 /s D
Gas mean residence time seconds 𝜃𝑔
Area m2 A
Liquid height m H Do not confuse with
Henry’s constant
Oxygen transfer rate g or mmol/l,h OTR
Volumetric oxygen g or mmol/l,h ro Also called OCR (Oxygen
consumption rate Consumption Rate)
Specific oxygen g/g cells,h or mmol/g qo Oxygen consumption by
consumption rate cells,h each cell
Surface tension N/m 𝛾 Force per unit length
Mean rising velocity of gas m/s V
bubble
Liquid film thickness m 𝛿
Mass flux by diffusion g/m2 ,s J
Total bubble area/total m2 /m3 A
liquid volume
The mass transfer m/s k Mostly used for the liquid
coefficient side, kL
The liquid side volumetric s-1 kL a Depends on stirring and
mass transfer coefficient aeration
Henry’s constant L*bar/g kH Used to convert
concentration to pressure
or vv
Henry’s law constant L*100/g H Expression for Henry’s
(derived) constant used when we
convert DOT in g/L to %
Molar volume m3 substance/mol Vm Volume occupied by 1 mol
substance of substance. Used to
convert volume to mol.
Dissolved oxygen tension % DOT Oxygen value in the liquid
DOT at calibration % DOT* Oxygen value in the liquid
at calibration in
equilibrium with the gas

9.2 The macroscopic mass balance of oxygen

Let us now consider the mass balances for oxygen as we formerly did for cells, substrate and product, to better
understand the important parameters and to be able to calculate the consumption. To estimate the overall
oxygen consumption we start with a figure describing the control volume over which we make the balance.
The liquid flow is ignored, since the amount of oxygen in the liquid is very much less than the amount in
the gas. We start thereafter from the common mass balance expression but now over the gas phase. The
macroscopic mass balance on oxygen is given by:
134 9 Aeration

Figure 9.4: Control volume for mass balancing of oxygen and carbon dioxide in a bioreactor. The oxygen in the liquid in- and outflow is disregarded
due to the comparatively low concentration

Definition 9.2.1 The macroscopic mass balance over the gas phase

𝑑(𝑦 𝑜𝑢𝑡 ∗ 𝑉)
= 𝑄 𝑖𝑛 ∗ 𝑦 𝑖𝑛 − 𝑄 𝑜𝑢𝑡 ∗ 𝑦 𝑜𝑢𝑡 ± 𝑟 𝑦 ∗ 𝑉 (9.6)
𝑑𝑡

In the mass balance, the volumetric oxygen consumption rate, 𝑟 𝑜 , is used and this unit is negative since
oxygen is consumed as all other substrates. As for the other rates so far discussed it is composed of the specific
oxygen uptake rate and the cell concentration. The volumetric consumption of oxygen, 𝑟 𝑜 , is so central and
important that it is also given a specific abbreviation “OCR” (Oxygen Consumption Rate) which equals:

𝑂𝐶𝑅 = 𝑟 𝑜2 = 𝑞 𝑜 ∗ 𝑐 𝑥 (9.7)

For the specific cell mass and the substrate consumption rates, 𝜇 and 𝑞 𝑠 , the Monod model has been used in
previous chapters to describe how the concentration of the substrate influences these parameters. For oxygen,
we may use the corresponding model meaning that:

𝑐 𝑜2
𝑞 𝑜 = 𝑞 𝑜,𝑚𝑎𝑥 (9.8)
𝐾 𝑠,𝑜 + 𝑐 𝑜2

Note however that “c𝑜2 ” now refers to the oxygen concentration and the saturation constant, Ks,o , corresponds
to the affinity for oxygen uptake. We have now gotten the expressions necessary for the consumption but
how can we describe the oxygen transfer from gas to liquid?

9.3 The molecular oxygen transfer rate, OTR

With respect to the oxygen transfer we realise that the oxygen introduced to the reactor from the sparger
must, once inside the reactor, be transported from the gas to the liquid in order to reach the cells. This phase
transfer is unique for this substrate (and carbon dioxide of course). We need therefore an expression for this
flux, which constitutes the oxygen transfer rate (abbreviated: OTR) to the liquid and preferably in a form
9.3 The molecular oxygen transfer rate, OTR 135

which couples it to important design parameters that we can also quite easily measure. We will, as we did for
the transfer of momentum, turn to a molecular transport model for this purpose.
The molecular balance for oxygen transfer is thus made over the gas-liquid interface where the molecular
transport takes place by diffusion rather than by convection. All diffusion processes where a mass component,
“i”, is transported a certain distance are described by Fick´s first law of diffusion which reads:

𝑑𝑐 𝑖
𝐽𝑖 = 𝐷 (9.9)
𝑑𝑥

This flux is often given a negative sign to show that transport takes place from high to low concentration (as
for momentum transfer), but for reasons of simplicity we here consider the quantity to be positive.
Fick´s law states in words that: the oxygen flux, J, over a specific distance (dx) is directly proportional to a
transfer coefficient over this surface, i.e. “D”, the diffusivity [m2 /s], multiplied by the concentration gradient
( 𝑑𝑐 𝑖 ). This gradient is the driving force for the transport (cf. momentum flux and the velocity gradient!). D/dx
has the unit [m/s] and the concentration is [g/m3 ]. The flux unit is thus [g/m2 ,s]. The transport process is
schematically outlined below.

Figure 9.5: Transport of oxygen from a gas bubble to the liquid where the cells are situated. The blue line indicates the drop in oxygen concentration
between the phases but indicates a homogeneous concentration within the gas bubble. The concentration in the liquid phase is here constant which in
reality might not be true

To reach the liquid is of course not the end of the story, oxygen must also to arrive as far as to penetrate
the cell membranes and reach the cytochromes for oxidation to water. This latter transport process will
not be further discussed in this context and is actually often disregarded in modelling. But it is easy to
understand that if cells grow in large flocks then those in the middle might even be subjected to anaerobic
conditions. This means that all cells in the population do not function in the same manner, i.e. they have
different physiology due to differences in the oxygenation. What this means for the production has to be
explored for the individual case.
Figure 9.5 is however too simplified for our purposes and does not include a clear boundary over which
we perform our balances. If we look at the process into more detail, we see that the oxygen must first be
transported inside the gas bubble, pass an outer boundary in the bubble called the “gas film” where we
believe the concentration is lower and move further through a liquid boundary “liquid film” to finally reach
the liquid bulk. This is drawn below.
This flux model is called the “Two-film theory” due to the postulation of the existence of the gas and liquid
films, or “interphases” where the concentration is different from the bulk of gas or liquid. It is thus considered
that the major drop in oxygen concentration take place over these films. The two-film model shows the flux
of oxygen from the gas into the liquid assuming that all the oxygen that leaves the gas enters the liquid which
136 9 Aeration

Figure 9.6: Schematic drawing of the two-film model showing the passage of oxygen from gas to liquid. The film thicknesses, or interphases, are
denoted, dx and dy, for the gas and liquid side, respectively

seems reasonable when there is a high consumption rate in the liquid bulk. The corresponding mathematical
expressions based on Fick´s law will be:

𝐷𝑔
𝐽𝑔 = (𝑝 − 𝑝 𝑔 ) (9.10)
𝑑𝑥

𝐷𝐿
𝐽𝐿 = ∗ (𝑐 𝑖 − 𝑐) (9.11)
𝑑𝑦

However, since it is well established that the liquid-side resistance is much larger than in the gas, the gas side
term is neglected, leaving the following expression:

𝐷𝐿
𝐽= ∗ (𝑐 𝑖 − 𝑐) (9.12)
𝑑𝑦

where J is the flux of oxygen over the film thickness dy and the driving force is the gradient where " 𝑐 𝑖 " is the
concentration in the film and "c" the concentration in the bulk. The term 𝐷𝐿 indicates that only the liquid
side diffusion is being considered.
The concentrations in the films are however unknown. To obtain a more useful expression, we realise that the
oxygen concentration, in the liquid-side film, is in equilibrium with the gas phase. We use this value is instead
and denote it “ 𝑐 ∗𝑜 2 ”, and use “ 𝛿 ” as an expression for the film thickness. The new expression according to the
reduced two-film theory is thus:

𝐷𝐿
𝐽𝑜 2 = ∗ (𝑐 ∗𝑜2 − 𝑐 𝑜2 ) (9.13)
𝛿

From this expression, it is evident that the oxygen flux can be increased if the film thickness 𝛿 can be reduced.
A way to achieve this is by a more vigorous agitation which leads to increased turbulence in the medium.
Graphically, the reduced flux model is shown in Figure 9.7.
Two different situations may appear that will influence the steepness of the oxygen concentration curve in the
liquid i.e. how rapidly the oxygen is consumed. This depends on whether it is the transfer or the consumption
of oxygen that is the faster process i.e. is limiting. If the consumption is faster then the transfer, oxygen will
rapidly disappear from the liquid in a “mass transfer limited” process. If the transfer is the faster process, the
process is “reaction rate limited” and we may always detect oxygen in the liquid bulk phase.
9.3 The molecular oxygen transfer rate, OTR 137

Figure 9.7: Oxygen transfer model based on the reduced two-film theory. C* is the concentration, in the liquid film, which is in equilibrium with the
gas phase, here considered to be homogeneous. Not drawn to scale!

The expression DL / 𝛿 is usually replaced by the expression: “the mass transfer coefficient”, denoted “kL ”,
which has the unit [m/s], and where the suffix L again shows that it is the liquid side coefficient that is more
important. Remember however that kL is strictly depending on the diffusivity and film thickness because
these are the parameters that can be used to enhance this value - a large value results in a high flux of oxygen,
J!
To get a better appreciation of what we call the oxygen transfer coefficient, kL , we suggest that this coefficient
represents a sum of the resistance to oxygen transfer from the gas to the liquid and we will see that the
transfer coefficient is actually inversely proportional to this sum. To get a better understanding of this, we
recall that the total resistance of a set of parallel resistances in an electrical circuit obey the Kirchoff´s law for
parallel resistors according to:

1 1 1 1 1
= + + + + ... (9.14)
𝑘 𝑘1 𝑘2 𝑘3 𝑘4

If we use this model also for our case of diffusion and replace the numbers 1-4 by the individual resistances
in the parallel gas/liquid bulk and interfaces, respectively, we see that if we add them up we get the total
resistance, 1/k. And this is actually the term that we have replaced by kL . This is pictured in Figure 9.8.

Figure 9.8: The sum of the parallel mass transfer resistances that make up the volumetric mass transfer coefficient kL
138 9 Aeration

We have however not yet arrived at the final flux expression for the OTR since we do not want to approximate
the flux over a distance but over the total bubble area of the liquid volume in the reactor. To get here, we
multiply both sides of the equation with the total bubble area per volume, called “a” which is:

𝑡𝑜𝑡 𝑎𝑙 𝑏𝑢𝑏𝑏𝑙𝑒 𝑎𝑟𝑒 𝑎 𝑚2


 
𝑎=
𝑡𝑜𝑡 𝑎𝑙 𝑙𝑖𝑞𝑢𝑖𝑑 𝑣𝑜𝑙𝑢𝑚𝑒 𝑚3

The revised Equation 9.13 will thus be:

𝐽𝑜2 ∗ 𝑎 = 𝑘 𝐿 ∗ 𝑎 ∗ (𝑐 ∗𝑜2 − 𝑐 𝑜2 ) (9.15)

From this it becomes evident that the flux is directly proportional to the total surface area of the bubbles,
something that we have only hinted before but now show mathematically.
The product of the flux times the total bubble area (J*a) is in bioprocess engineering called the volumetric
Oxygen Transfer Rate, OTR and the final expression for this parameter where the unit for the OTR is
[g/m3 ,h] or [mmol/l,h] is now given by:

𝑂𝑇𝑅 = 𝑘 𝐿 ∗ 𝑎(𝑐 ∗𝑜2 − 𝑐 𝑜2 ) (9.16)

The coefficient “kL a” is one of the most important parameters describing oxygen transfer in aerobic
bioprocesses and is called the volumetric mass transfer coefficient, and is usually expressed in [h-1 ]. For a
certain bioreactor and medium, this parameter can be used to describe the efficiency for oxygen transfer.
When a new bioreactor is purchased, this parameter is declared by the retailer to characterize the oxygen
transfer capacity. The methods for the estimation of this parameter (which are described further below) and
the actual value should be carefully studied before a purchase is made. The kL a is dependent on several
parameters but for a given reactor/process, the gas flow and the stirring rate are the most important.
The amount of oxygen in the liquid phase is measured in most cultivations i.e. most often during process
development and quite often also in production. We use an oxygen electrode for this purpose and such
electrodes can be built as based on some different concepts usually on galvanic or polarographic principles.
Recent techniques also include optical methods. It should be carefully noted that the response of the electrode
is proportional to the partial pressure of oxygen in the reactor. To get the concentration we use Henry´s law
which states that at constant temperature the concentration of a given gas dissolved in a given type and
volume of liquid is directly proportional to the partial pressure of that gas in equilibrium with that liquid,
according to:

𝑝 𝑜2 = 𝑘 𝐻 ∗ 𝑐 𝑜2 (9.17)

where kH is Henry´s constant which can be found in standard tables for different temperatures.
To follow rapid changes of oxygen in the liquid phase, it is important that the response time of the electrode
is fast (remember how fast oxygen is depleted after a stop in the aeration!) otherwise we will just measure the
response time of the electrode. The electrode is calibrated before the beginning of the cultivation and the
signal allowed some time to stabilize after air is introduced into the reactor. When the value is in equilibrium
with air, it is adjusted to a value that we arbitrary set to 100% and this is this value that we call: DOT*
(Dissolved Oxygen Tension). During cultivation, this DOT-value decreases as the cells consume the oxygen as
we have earlier seen for the different cultivation concepts. For simplicity, and to relate to the electrode value,
the OTR-equation is usually written as:

𝑂𝑇𝑅 = 𝑘 𝐿 𝑎 ∗ (𝐷𝑂𝑇 ∗ − 𝐷𝑂𝑇) (9.18)


9.3 The molecular oxygen transfer rate, OTR 139

The DOT*, which corresponds to “ 𝑐 ∗𝑜2 ” is = 100% provided the gas phase is not depleted of oxygen, and the
definition is:

𝑐 𝑔𝑎𝑠𝑜𝑢𝑡 𝑝 𝑎𝑐𝑡𝑢𝑎𝑙
𝐷𝑂𝑇 ∗ = 100 ∗ ∗ (9.19)
𝑐 𝑔𝑠𝑐𝑎𝑙 𝑝 𝑐𝑎𝑙

This relates the calibrated values to the actual values at the cultivation start i.e. before inoculation. Since
the concentration, 𝑐 𝑔𝑎𝑠𝑜𝑢𝑡 , in the outlet gas from the bioreactor before cultivation starts is often similar to
the value at calibration and if the total pressure in the reactor does not change during cultivation, the initial
calibration value, DOT*, does not generally deviate from 100%. This is important to recognise since the actual
DOT value is related to this start-value and, if this is changed, e.g. if the total pressure in the reactor is
changed, the scale of the DOT measurement is changed.
The value of DOT* will not change over the whole liquid bulk of the reactor if the gas phase concentration
does not change. This is however only valid for small reactors. In a large airlift reactor, this value will decrease
progressively as the top of the reactor is approached and as the total pressure is reduced when we approach
the surface. For a more reliable estimation, it is possible to replace this value on the basis of some physical
property such as (i) the ingoing or (ii) the outgoing value for the whole reactor, (iii) a linearly decrease in the
value in proportion to the height in the bioreactor or (iv) by using a compensation according to the calculated
hydrostatic pressure.
We are now ready to summarise the mass balance based on the molecular transfer of oxygen from gas to
liquid (compare to the balance over the whole reactor in the beginning of the chapter!) where we now get the
concentration in the liquid as:

𝑑𝑐 𝑂2 ∗
= 𝑂𝑇𝑅 − 𝑂𝐶𝑅 = 𝑘 𝐿 𝑎 ∗ (𝑐 𝑂 − 𝑐 𝑂2 ) − 𝑟 𝑜2 (9.20)
𝑑𝑡 2

Some common values for the oxygen transfer parameters are given in Table 9.2 for the cultivation of some
different organisms to exemplify the magnitudes.

Table 9.2: Examples of oxygen transfer parameter in some common processes

Process OTR (mmol/l,h) kL a


Waste water 1.6 - 4.8 10 - 20
Antibiotics 18.8 70 - 250
Yeast 312 700

The volumetric mass transfer coefficient has been extensively studied in different reactors and for guidance,
empirical correlations have been established. In such correlations kL a is generally correlated to the power
input/volume and the aeration rate which we know will radically affect the value. Typical correlations are:

𝑘 𝐿 𝑎 = 2.6 ∗ 10−2 ∗ (𝑃/𝑉)0.4 ∗ 𝑣 0𝑠 .5 𝑓 𝑜𝑟 𝑤𝑎𝑡𝑒𝑟 (9.21)

𝑘 𝐿 𝑎 = 2.0 ∗ 10−3 ∗ (𝑃/𝑉)0.7 ∗ 𝑣 0𝑠 .2 𝑓 𝑜𝑟 𝑤𝑎𝑡𝑒𝑟 (9.22)

These expressions are based on a correlation suggested by van´t Riet in 1979 and show that the kL a is strongly
affected by the mixing (P/V) but also the rate of gassing, expressed as superficial gas velocity. We see also
that we need different expressions depending on the medium used. One question at this point might be:
which of the two parameters P/V or the gas flow is most important?
Let us study the graph in Figure 9.9. In this case we see that the measured values correlate well to the
empirical equation and we can also see that both parameters are almost equally important for efficient mass
transfer.
140 9 Aeration

Figure 9.9: Dependence of the volumetric gas transfer coefficient on the power/volume (P/V) and the gas flow expressed as the superficial gas velocity,
here denoted wsg. The measured points (open ring) are fitted to the equation. Courtesy by Prof A Lübbert, Univ. of Halle

If we look at mass transfer data from the two common types of bioreactors: the STR and the bubble column
we will see how different levels of the superficial gas velocity (or the energy dissipation rate) and the medium
properties affect the kL a-value.
It is evident that only at very high linear gas flows does the bubble column have an oxygen transfer comparable
to that in the STR. The medium coalescence obviously has a tremendous effect on the kL a. Further, we see
that kL a is mostly affected at low energy dissipation rates in particular for coalescent media.

Figure 9.10: The volumetric mass transfer coefficient as a function of the superficial gas velocity (which is proportional to the energy dissipation, 𝜀)
for two reactor types and two different media

Foaming of the medium in bioreactors is a serious problem in many cultivations and if foaming suddenly
starts it can empty a bioreactor in a very short time unless counter-measures are taken. It should perhaps be
noted already here that the mechanism for foaming should not be confused with the gas holdup that was
shown in an earlier figure since the mechanisms are different. Increased foaming is often seen when the gas
flow is increased and the foam collects generally on top of the liquid broth.
Surface-active compounds like proteins reduce the interfacial tension (see surface tension) between gas and
liquid leading to smaller bubbles (foaming) and this can sometimes be observed as an increase in oxygen
transfer during foaming. In order to overcome the problem, antifoam chemicals are often added to the
cultivation. These are often surface active compounds that accumulate at all interfaces including the interface
between gas and liquid. This increases the film thickness and leads to a radically reduced kL a-value. When
antifoam is added, the DOT value is often immediately reduced to zero, which is of course undesirable but
can often not be avoided.
The installation of stirrer-like devices under the top of the reactor lid, in the gas phase, will also reduce the
9.4 Spargers 141

Figure 9.11: A foam head at the top of the liquid in a bioreactor

foaming, in this case by mechanical means. These mechanical foam breakers rotate and the pressure created,
and the resulting shear forces, reduce the foam. These can also be coupled to additions of a chemical to
reduce the foam if the mechanical capacity is not sufficient.

9.4 Spargers

The design of the spargers for the introduction of air into the reactor depends on the reactor type. Ring
spargers are frequently used for STR´s, but the sparger used in an air-lift reactor can be of various designs
and usually covers most of the bottom of the reactor to get a full gas dispersion of the reactor cross section.

Figure 9.12: Different sparger designs for different reactor types. Left: ring sparger positioned under reactor stirrer. Right: top view of the sparger
covering the bottom of an airlift reactor

The material of the sparger and the holes for air passage affect the bubble size and thus the OTR as we have
summarised before. An example of the effect of the bubble size created by different spargers is shown in
Figure 9.13.
Here, drilled holes in a pipe are compared to transport over a porous material where the latter will lead to
formation of considerably smaller bubbles. The graph to the left shows the time taken to degasify a bioreactor
by a change in the inlet flow from oxygen to nitrogen. The more rapid decrease in the oxygen concentration
142 9 Aeration

Figure 9.13: Effect of the sparger material on the oxygen transfer rate. Left: gassing out of dissolved oxygen with nitrogen. It is obvious that the
smaller holes in the porous material leads to a much faster gassing-out than the drilled pipe. Right: bubble swarms from porous sparger material

makes it evident that the porous material leads to a much higher transfer rate than the drilled pipes (see the
slope). The efficiency of the porous material is illustrated in the right part of the figure where large swarms
of very small bubbles are created.

9.5 Methods to estimate the kL a

Three principally different methods are available for the determination of the volumetric mass transfer
coefficient and they are applicable under quite different conditions where the methods differ mainly in their
ability to accommodate cells or not. The methods are:

I The gas balance method (with and without sulphite oxidation)


I The gassing out method
I The dynamic method (using an interruption of the aeration for gassing out by use of the oxygen
consumption of cells)

The gas balance method

Consider a fully aerated and stirred bioreactor. We will start to derive the method by looking at the equation
describing the gas/liquid interface transport,

𝑑𝑐 𝑜2
= 𝑘 𝐿 𝑎(𝑐 ∗𝑜2 − 𝑐 𝑜2 ) − 𝑂𝐶𝑅 (9.23)
𝑑𝑡

In the first attempt to use this balance, we may assume that it is possible to reach a situation where the oxygen
concentration in the reactor is zero, 𝑐 𝑜2 = 0, and consequently d 𝑐 𝑜2 /dt=0 although we add large amounts
with the air. In this case all oxygen that is transferred is immediately consumed i.e.:

𝑂𝑇𝑅 = 𝑂𝐶𝑅

We may then the use the reactor balance to estimate the OCR and we start from the balance equation:

𝑑(𝑐 𝑜2 ,𝑜𝑢𝑡 ∗ 𝑉)
= 𝑄 𝑖𝑛 ∗ 𝑐 𝑜2 ,𝑖𝑛 − 𝑄 𝑜𝑢𝑡 ∗ 𝑐 𝑜2 ,𝑜𝑢𝑡 − 𝑟 𝑜2 ∗ 𝑉 (9.24)
𝑑𝑡

Where there is no change in the outgas with time ( 𝑑𝑐 𝑜2 ,𝑜𝑢𝑡 /𝑑𝑡 = 0) the expression for the consumption is:

𝑄 𝑖𝑛 ∗ 𝑐 𝑜2 ,𝑖𝑛 − 𝑄 𝑜𝑢𝑡 ∗ 𝑐 𝑜2 ,𝑜𝑢𝑡


𝑟𝑜2 = = 𝑂𝐶𝑅 (9.25)
𝑉
9.5 Methods to estimate the kL a 143

By insertion in the gas/liquid transport equation above we get (provided there is no oxygen accumulation in
the liquid as we stated above):

𝑄 𝑖𝑛 ∗ 𝑐 𝑜2 ,𝑖𝑛 − 𝑄 𝑜𝑢𝑡 ∗ 𝑐 𝑜2 ,𝑜𝑢𝑡


𝑘 𝐿 𝑎 ∗ 𝑐 ∗𝑜2 = (9.26)
𝑉

We see that if the concentration in the outlet gas can be determined, we can thus easily estimate the kL a. This
expression refer to the maximum oxygen transfer capacity [mmol/L,h] which will take place when the driving
force is the greatest i.e. when 𝑐 𝑜2 = 0. Note that this expression differs with respect to the “volumetric oxygen
transfer coefficient, kL a”, which has the unit [h-1 ], since it includes the oxygen equilibrium concentration,
𝑐 ∗𝑜2 . To obtain the kL a we can of course divide the resulting expression with this factor. Many retailers of
bioreactors choose however to report the 𝑘 𝐿 𝑎 ∗ 𝑐 ∗𝑜2 . The unit indicates which value that is actually given and
this should be carefully acknowledged.
How can we get to a situation where 𝑐 𝑜2 = 0? A common method is to chemically reduce the oxygen
concentration in what is called “the sulphite oxidation method”. Sulphite is here added in large amounts
to the reactor medium and is oxidized to sulphate in a fast reaction consuming all the oxygen. This means
that the system is mass transfer controlled, which is the situation needed to get the parameter we desire to
measure. The reaction is:

1
𝑁 𝑎 2 𝑆𝑂3 + 𝑂 2 → 𝑁 𝑎2 𝑆𝑂 4
2

where a small amount of copper sulphate is used as a catalyst. The equilibrium is driven far to the right so
that hardly any oxygen is present. The value of 𝑘 𝐿 𝑎 ∗ 𝑐 ∗𝑜2 is then obtained from the ingoing gas concentration
(≈21%) and the outgas value which is often available. The values of the gas in- and outflow are equal since
we have no gas production in this case. We will repeat this experiment at the different stirring or aeration
rates in which we are interested. This technique is of course not used with growing cells since they will not
withstand the prevailing conditions.
We note also that the consumption rate of oxygen can be calculated from the reaction rate of the sulphite
oxidation from titration of unreacted iodine with starch as indicator. This gives also the value of the specific
mass transfer coefficient times c*, but without the need for outgas analysis.
In the second attempt to use the balance equations, when we cannot, or do not want to arrive at a situation
where c = 0, we use the oxygen electrode to get a DOT-reading as we change the stirring and the aeration
rates. We start again from:

𝑄 𝑖𝑛 ∗ 𝑐 𝑜2 ,𝑖𝑛 − 𝑄 𝑜𝑢𝑡 ∗ 𝑐 𝑜2 ,𝑜𝑢𝑡


𝑘 𝐿 𝑎(𝑐 ∗𝑜2 − 𝑐 𝑜2 ) = (9.27)
𝑉

If we multiply both sides by 𝑐 ∗𝑜2 /(𝑐 ∗𝑜2 − 𝑐 𝑜2 ) and convert the concentrations to DOT i.e. to DOT*/(DOT*-DOT)
by use of the conversion factor H according to:

𝑝 𝑜2
𝑐 𝑜2 = (9.28)
𝑘𝐻
and
𝑝 𝑜2
𝐷𝑂𝑇 = 100 ∗ (9.29)
𝑝𝑜
2 ,𝑐𝑎𝑙

Which when combined gives:


𝐷𝑂𝑇 ∗ 𝑝 𝑜 𝐷𝑂𝑇
2 ,𝑐𝑎𝑙
𝑐 𝑜2 = = (9.30)
𝑘 𝐻 ∗ 100 𝐻
144 9 Aeration

and we obtain:
𝑄 𝑖𝑛 ∗ 𝑐 𝑜2 ,𝑖𝑛 − 𝑄 𝑜𝑢𝑡 ∗ 𝑐 𝑜2 ,𝑜𝑢𝑡 𝐷𝑂𝑇 ∗
𝑘 𝐿 𝑎 ∗ 𝑐 ∗𝑜2 = ∗ (9.31)
𝑉 𝐷𝑂𝑇 ∗ − 𝐷𝑂𝑇

This is a method that can also be used on-line during a cultivation to estimate the oxygen transfer capacity. It
should be realised however, that if the DOT values are close to the maximum (e.g. at a low consumption rate),
quite large errors appear due to the division of two similar numbers. Another drawback of this method is
that we rely on the accurate measurement of the oxygen electrode. Since these electrodes often have long
response times, there is a risk that it is only these response times that are measured. This corresponds to the
use of a reaction-rate-controlled process rather than one that is transfer-rate controlled which we in this case
is interested in.
Note also that the values that are given from the outgas measurements are given in volume% while the oxygen
transfer capacity is given in [mmol/L,h]. By using the molar volume of oxygen at the given temperature (Vm
[L/mol]) the correct units are achieved and the final equation will be:

𝑄 𝑖𝑛 ∗ 𝑜𝑥 𝑦 %𝑖𝑛 − 𝑄 𝑜𝑢𝑡 ∗ 𝑜𝑥 𝑦 %𝑜𝑢𝑡 𝐷𝑂𝑇 ∗


 
1000
𝑘𝐿 𝑎 ∗ 𝑐 ∗𝑜2 = ∗ ∗ (9.32)
100 ∗ 𝑉 𝐷𝑂𝑇 − 𝐷𝑂𝑇 𝑉𝑚

Gassing out with nitrogen

When the gassing out method is used, the starting point is once again the gas balance over the gas-liquid
interface:

𝑑𝑐 𝑜2
= 𝑘 𝐿 𝑎(𝑐 ∗𝑜2 − 𝑐 𝑜2 ) − 𝑟 𝑜2 (9.33)
𝑑𝑡

but since an oxygen electrode is used also here, the preferred form is based on the DOT-value (that is actually
measured) and we use the conversion factor DOT=c*H as before:

𝑑𝐷𝑂𝑇
= 𝑘 𝐿 𝑎(𝐷𝑂𝑇 ∗ − 𝐷𝑂𝑇) − 𝑟 𝑜2 ∗ 𝐻 (9.34)
𝑑𝑡

The actual experiment proceeds as follows. Consider a medium, in this case without cells that has been
aerated to equilibrium with oxygen in the air. The experiment is then started by introducing nitrogen gas
through the sparger which successively leads to a decreasing DOT value that eventually approaches zero
(gassing out). After some time, the aeration is restarted and the oxygen signal from the electrode is again
recorded. The DOT curve appears as follows:
To calculate the kL a by this method we realise that since we do not use any cells, this means that the volumetric
oxygen consumption, ro , is equal to zero and the equation, given further above, is reduced to:

𝑑𝐷𝑂𝑇
= 𝑘 𝐿 𝑎(𝐷𝑂𝑇 ∗ − 𝐷𝑂𝑇) (9.35)
𝑑𝑡

This can be integrated if we first make the rearrangement:

1
∗ 𝑑𝐷𝑂𝑇 = 𝑘 𝐿 𝑎 ∗ 𝑑𝑡 (9.36)
𝐷𝑂𝑇 ∗ − 𝐷𝑂𝑇

Integration takes place from t=0, which is the start of the experiment i.e. when air is again introduced and
where DOT=DOT*= 0 (all oxygen is gone), to t=t and the final DOT value. Note that the integration requires a
9.5 Methods to estimate the kL a 145

Figure 9.14: Gassing out method for kL a estimation without cells. Measurements of DOT start at the onset of air when DOT is close to zero

substitution since we basically wish to perform an integration of the character: 1/(1-x) dx. Substitution of
(DOT*-DOT) results in:

𝑙𝑛(𝐷𝑂𝑇 ∗ − 𝐷𝑂𝑇) = −𝑘 𝐿 𝑎 ∗ 𝑡 (9.37)

This can be recognised as a straight line with a negative slope. This will not intercept the y-axis since ln[0] is
not defined. The negative slope gives the value of kL a.

Figure 9.15: Determination of the volumetric oxygen transfer coefficient by the “gassing out” method. The oxygen mass transfer coefficient is given
from the slope of the curve ln(DOT*-DOT)=-kL a*t

The dynamic method

The dynamic method can be used in the presence of cells and the method is very similar to the gassing
out method. Here however, the gassing out is driven by the consumption of oxygen by the cells when the
air to the bioreactor is interrupted rather than by sparging with nitrogen. This method is also based on
measurements relying on the accuracy of the oxygen electrode with the aforementioned drawbacks.
It is important that the experiment does not take too long a time since it builds on the assumption that the
oxygen consumption rate is not radically changed during this time and thus that the start and the final
DOT values are almost equal. We need however to perform the experiment at a sufficiently high cell mass
concentration to support a rapid consumption of the oxygen. The starting point is again the gas/liquid mass
146 9 Aeration

balance on oxygen. Since we have cells present, we cannot remove the consumption rate from the mass
balance and the equations will be slightly more complex.

𝑑𝐷𝑂𝑇
= 𝑘 𝐿 𝑎(𝐷𝑂𝑇 ∗ − 𝐷𝑂𝑇) − 𝑟 𝑜2 ∗ 𝐻 (9.38)
𝑑𝑡

At a specific point in the cultivation, and at a certain arbitrary DOT value, the air supply is turned off. The
cells will continue to consume the oxygen and the DOT value is continuously reduced. This is plotted in the
left-hand part of the graph below. This process is continued until DOT is reduced to a considerable amount.
To protect the cells the value is however not allowed to be too low.
During this first part of the experiment, the air is shut-off and there is consequently no transfer of oxygen to
the reactor and the change in the DOT is solely due to consumption and the mass balance is reduced to:

𝑑𝐷𝑂𝑇
= −𝑟 𝑜2 ∗ 𝐻 (9.39)
𝑑𝑡

Figure 9.16: Dynamic method for kL a determination with cells present

From this equation, i.e. the slope of the curve, we will obtain the value for the consumption of oxygen. The
integration gives:

𝐷𝑂𝑇2 − 𝐷𝑂𝑇1
−𝑟 𝑜2 ∗ 𝐻 = (9.40)
𝑡2 − 𝑡1

When sufficient data are logged from this first part, the oxygen supply is again turned on and the DOT value
then increases. Here we must to use the whole oxygen mass balance since there is a simultaneous transfer
and consumption of oxygen. If the experiment is rapid enough, the value for 𝑟 𝑜 ∗ 𝐻 from the first part of
the experiment can be used and inserted into the equation. Rewriting the equation gives the straight-line
relationship:

𝑑𝐷𝑂𝑇
+ 𝑟 𝑜2 ∗ 𝐻 = 𝑘 𝐿 𝑎 ∗ (𝐷𝑂𝑇 ∗ − 𝐷𝑂𝑇) (9.41)
𝑑𝑡

The value of dDOT/dt is given by the slope of the curve that is created at the particular time points (see
Figure 9.16). When this is determined, 𝑑𝐷𝑂𝑇/𝑑𝑡 + 𝑟𝑂2 ∗ 𝐻 can be plotted versus DOT*-DOT (where the
DOT*-value is 100% and DOT the actually considered point value). The slope gives the value of the kL a.
9.6 Mixing and aeration 147

9.6 Mixing and aeration

Most cultivation processes are both mixed and aerated and, as we have seen, these two processes heavily
depend upon each other. This was empirically described by Michel Miller already in 1962 according to:

" # 0.44
𝑃02 ∗ 𝑁 𝐷𝑖3
𝑃 𝑔 = 0.832 (9.42)
𝑄 0.56

and from this equation, we get the first hint that a high airflow will reduce the gassed power (Pg ) input
needed. To study this correlation we may also here use a dimensionless number called the aeration number,
Na , which relates the air velocity to the tip speed of the impeller. This number is (for e.g. 𝐷𝑡 = 2𝐷𝑖 ) defined
as:

𝑄
𝜋𝐷 2
𝑡
𝑠𝑢𝑝𝑒𝑟 𝑓 𝑖𝑐𝑖𝑎𝑙 𝑔𝑎𝑠 𝑣𝑒 𝑙𝑜𝑐𝑖𝑡 𝑦 4 𝑄
𝑁𝑎 = = = (9.43)
𝑖𝑚𝑝𝑒 𝑙𝑙𝑒𝑟 𝑡𝑖𝑝 𝑠𝑝𝑒 𝑒 𝑑 𝜋 ∗ 𝑁 𝐷𝑖 𝑁 𝐷𝑖3

If we look at the power drawn at different air flow rates in a reactor (Pg ) in comparison to the non-gassed
situation (Po ), with increased aeration number, we can see that the power drawn is considerably reduced
when air is introduced:

Figure 9.17: Gassed power consumption as a quotient of ungassed power and a function of the aeration number

Sometimes, when we perform a cultivation we may look at the DOT electrode and find that the value is “too
low” and that we need to increase the oxygen transfer rate (OTR). We try to do this by increasing the airflow,
i.e. the VVM, but nothing happens! The reverse may also be true - an increase of the RPM does not increase
the DOT. How can this be explained?
To understand this, we have to recall the models we have derived which show the parameters that control the
oxygen value, DOT, and the oxygen transfer coefficient of the reactor. If there is no increase in DOT when the
airflow is increased at constant RPM, then the stirring is probably too low and unable to distribute more
air and/or divide the air into smaller bubbles. A plateau of the OTR has been reached. This can be seen in
the Figure 9.18 at constant RPM (orange). In the opposite situation, when the airflow is constant, increased
stirring gives finally no benefit since there is no more air to distribute and also here a plateau is reached
(constant VVM at two different levels, in green).
148 9 Aeration

Figure 9.18: Correlation between the stirring speed, RMP and the aeration rate, VVM

The situation is further illustrated in Figure 9.19. Moving to the right in the figure shows how an increased
stirring leads to a better gas dispersion with a constant airflow. Moving towards the left may be interpreted
as moving to a higher gas flow with constant stirring. To the far left, the agitation is insufficient to distribute
the gas, a situation called “flooding” of the agitator, whereas at the far right, the gas is fully dispersed.

Figure 9.19: Gas dispersion by an agitator. From left to right: increasing agitation with constant airflow. From right to left: increased aeration with
constant agitation. The left-hand reactor shows that the impeller is flooded, the middle reactors different cases of loading and the right-hand reactor
shows a full dispersion of the gas

In large mixed and aerated reactors, the oxygen transfer performance is often insufficient, since there is just
not enough capacity available. The goal of perfect homogeneity has therefore to be rejected.

Figure 9.20: CFD simulation of large-scale reactor performance. Gas holdup (left), kL a distribution (middle) and the distribution of bubble sizes
(right). Red, at the top of the legend to the far left, is the highest value and blue is consequently the lowest

The CFD simulation in Figure 9.20 shows the common, and perhaps unavoidable, situation in a large reactor
9.6 Mixing and aeration 149

with three agitators positioned at some distance on top of each other. The figure shows the important
parameters controlling the oxygenation: the gas entrapment (gas hold-up), the bubble size distribution and
the oxygen transfer coefficient. The gas hold-up, shown to the left, shows that the gas is largely trapped close
to the impellers (in clear red) where the bubble sizes (the graph to the right) are particularly small (blue).
At some distance from the impellers little gas is present and the bubbles far from the impeller are very large
particularly at the top. Consequently, and in accordance with our theoretical discussion, the oxygen transfer
coefficient, kLa (middle graph), is highest close to the impellers (red) but rapidly decreases outside this
region, probably because the energy dissipation is, as stated before, greatest in the stirrer region.
What are then the limitations to the mixing and aeration design?
On the one hand, if all the mixing and aeration capacity available is utilised, this is accompanied by a
considerable cost which is not affordable for a number of processes - in particular where the product is
comparatively cheap. If the mixing is reduced and the aeration is maximised, foaming problems may arise.
This is due to the presence of small amounts of protein in the medium which are accumulated because of
the ever present cell lysis. Even if we are very careful with our treatment of the cells, there will always be a
certain disruption, basically due to cell dependent autolysis mechanisms but also due to harmful elements of
the physical/chemical environment.
The opposite procedure, i.e. a low aeration and maximised mixing, may lead to increased cell damage but
more particularly to a poor ventilation of the carbon dioxide. Too low a mixing and aeration altogether may
of course lead to too low an oxygen level. The parameters need therefore to be optimised for each process.

Figure 9.21: Limitations of high and low aeration and mixing

You might also like