THE TEMPERATURE LIMITED FED BATCH TECHNIQUE FOR10943/FULLTEXT01.pdf · THE TLFB TECHNIQUE FOR CONTROL OF E. COLI CULTURES 1 INTRODUCTION Biotechnology has been a major participant

THE TEMPERATURE-LIMITED FED-BATCH TECHNIQUE FOR

CONTROL OF ESCHERICHIA COLI CULTURES

MARIE SVENSSON

M.SC.

ROYAL INSTITUTE OF TECHNOLOGY

STOCKHOLM 2006

Marie Svensson

Stockholm, 2006

ISBN 91-7178-473- X

School of Biotechnology

Royal Institute of technology

106 91 STOCKHOLM

SWEDEN

ABSTRACT

The objective of this study was to investigate the physiology and productivity in Escherichia coli cultures with emphasis on the temperature-limited fed-batch (TLFB) culture. The TLFB technique controls the oxygen consumption rate of the growing culture by a gradually declining temperature from 37-35 °C down to 20-18 °C. The temperature regulated the DOT around a set-point (30 % air sat.), and all nutrients were in excess. Glucose was fed into the culture continuously, however, high acetate formation was avoided by keeping the glucose at a low, yet excessive, concentration. The biomass productivity was approximately the same in TLFB as in glucose-limited fed-batch (GLFB) cultures, since the specific growth rate and the oxygen consumption rate are limited by the oxygen transfer capacity of the reactor in both techniques. High concentrations of endotoxins were found in the medium of E. coli fed-batch cultures at low specific growth rates (below 0.1 h-1) and severe glucose limitation. In this thesis the TLFB technique was found to suppress the endotoxin release even at low specific growth rates. The repressed release of endotoxins in TLFB cultures was due to the high availability of glucose and not to the low growth rate or the lower temperature. The conclusion was drawn from comparing with the GLFB technique performed at 20 °C, which resulted in high endotoxin release. Extensive release of endotoxin, accompanied with high concentrations of soluble proteins was found in a TLFB culture exposed to a higher energy dissipation rate, 16 kW m-3, instead of 2 kW m-3, due to a higher stirrer speed (1000 instead of 500 rpm). The hypothesis that this is a result of mechanical stress is discussed in context with the common view that cells like E. coli, which are smaller than the Kolmogoroff’s microscale of turbulence, should not be influenced by the turbulence. TLFB cultured cells exhibited more stable cytoplasmic membranes when treated with osmotic shock as compared to the GLFB cultured cells. The concentrations of DNA and soluble proteins in the periplasmic extracts from the TLFB cultured cells were lower than from GLFB cultured cells. In addition, the specific productivity of periplasmic β-lactamase was higher in the TLFB cultures, suggesting that this technique could be an alternative for protein production. The solubility of a partially aggregated recombinant protein increased in the TLFB compared to the GLFB cultures. However some time after induction, in spite of the gradually declining temperature, the soluble fraction decreased. For obtaining better understanding of the performance of the process and for identifying critical parameters, a mathematical model was developed based on the growth, energy and overflow metabolism at non-limiting nutrient conditions. The temperature-dependent maximum specific glucose and oxygen uptake rates were determined in pH-auxostat cultures for temperatures ranging from 18 to 37 °C. A dynamic simulation model of the TLFB technique was developed and the results were compared to experimental data. The simulation program was also used for sensitivity analysis of some physiological and process parameters to study the impact on biomass concentration and temperature profiles. An effect on the biomass concentration profile but not on the temperature profile was observed when changing the oxygen transfer coefficient. If the maximum specific glucose uptake rate was altered, or if the glucose concentration was permitted to assume other values, the temperature profile but not the biomass concentration profile was influenced. Cell death affected both the biomass concentration profile and the temperature profile.

Keywords: endotoxin, Escherichia coli, fed-batch technique, glucose-limitation, mathematical model, mechanical stress, metabolic rates, osmotic shock, outer membrane, overflow metabolism, pH-auxostat, sensitivity analysis, simulation, stress response, temperature-dependence, temperature-limitation.

This book is especially dedicated to you…

Read it! (Yes, you.)

No…no no no NO! Don’t shut the book, don’t shut the book. DON’T…

LIST OF PUBLICATIONS

This thesis is based on the following publications, which will be referred to in the text with

their Roman numerals.

I. Control of Endotoxin release in Escherichia coli fed-batch cultures. Svensson M,

Han L, Silfversparre G, Häggström L, Enfors S-O. Bioproc. Biosyst. Eng. (2005) 27:91-

98.

II. Osmotic stability of the cell membrane of Escherichia coli from a temperature-

limited fed-batch process. Svensson M, Svensson I, Enfors S-O. Appl Microbiol

Biotechnol. (2005) 67(3):345-50.

III. Modeling of temperature-dependent growth of Escherichia coli in a pH-auxostat

and a temperature-limited fed-batch culture. Svensson M and Enfors S-O. (2006)

Submitted.

IV. Impact of stirring on the endotoxin release from Escherichia coli. Svensson M and

Enfors S-O. (2006) Manuscript.

In addition, previously unpublished results are included.

Additional publication not included in thesis

Dual substrate cultivation. Silfversparre G, Enfors S-O, Svensson M (2002) PCT No:

WO0246371.

TABLE OF CONTENTS

INTRODUCTION ........................................................................................................................................................ 1

OBJECTIVES ................................................................................................................................................................ 3

GROWTH CONTROL TECHNIQUES ................................................................................................................... 5

GLUCOSE LIMITATION ................................................................................................................................................ 7 TEMPERATURE LIMITATION (I) .................................................................................................................................. 7 PH-AUXOSTAT (III)..................................................................................................................................................... 9

PHYSIOLOGICAL RESPONSES........................................................................................................................... 13

STRESS RESPONSES ................................................................................................................................................... 13 THE E. COLI CELL AND OUTER MEMBRANES............................................................................................................ 15 ENDOTOXIN RELEASE (I) .......................................................................................................................................... 17 THE INFLUENCE OF STIRRING ON ENDOTOXIN RELEASE (IV) ................................................................................. 19 TEMPERATURE-DEPENDENT CYTOPLASMIC MEMBRANE STABILITY (II) ................................................................ 20

PROTEIN PRODUCTION ....................................................................................................................................... 23

PRODUCTION RATE AND PROTEIN SOLUBILITY (II) ................................................................................................. 25

MODELING THE TEMPERATURE-DEPENDENT RATES........................................................................... 29

THE OXIDATIVE AND OVERFLOW PATHWAYS.......................................................................................................... 29 MATHEMATICAL MODEL (III) .................................................................................................................................. 31 TEMPERATURE-DEPENDENT METABOLIC RATES (III) ............................................................................................. 33 SENSITIVITY ANALYSIS PERFORMED BY SIMULATIONS (IV)................................................................................... 36

CONCLUDING REMARKS .................................................................................................................................... 41

NOMENCLATURE ................................................................................................................................................... 43

ACKNOWLEDGEMENT......................................................................................................................................... 45

REFERENCES............................................................................................................................................................ 47

THE TLFB TECHNIQUE FOR CONTROL OF E. COLI CULTURES

1

INTRODUCTION

Biotechnology has been a major participant in the global industry ever since the development

of the recombinant DNA technology in the 1970’s. The number of products arising from

modern biotechnology, e.g. recombinant proteins produced from hosts with foreign gene

inserts, is large and continues to grow. The largest market shares are in the pharmaceuticals,

pulp and paper, and food and agriculture sectors (Bull et al. 2000). Over 60 % of the produced

enzymes used by the food, detergent and starch-processing industries are recombinant. About

eighty biotechnology drugs (recombinant protein drugs, vaccines and monoclonal antibodies)

were available on the market in the year of 2000 and even more were in clinical trials.

Examples of recombinant therapeutic proteins are erythropoietin (EPO), human growth

hormone (hGH), and human insulin (Demain 2000).

Among the microorganisms used in the biotechnology industry, Escherichia coli is one of the

best characterized both genetically and physiologically. It is frequently used because it is a

fast-growing organism; able to reach high cell-densities and growing on cheap, defined

media. The cultivation process is performed in the enclosed environment of a bioreactor in

one of the three main modes of cultivation; batch, fed-batch or continuous cultivation (fig 1).

In batch cultures, (fig 1, left), the essential substrates are included prior to inoculation,

however pH-titrating agents and oxygen are added to the culture. The growth is exponential

until some essential substrate becomes growth rate limiting, or until the culture is inhibited by

e.g. accumulated acetate, a by-product formed in the so-called overflow metabolism of E.

coli. The fed-batch culture (fig 1, center) is the most frequently used cultivation mode. The

culture is continuously fed with one or several essential substrate components in addition to

the titrating agent and the oxygen supply. Typically in fed-batch cultures, the substrate

concentration in the feed is very high to minimize the dilution of the culture. Furthermore, the

feeding rate is lower than the maximum substrate uptake rate, thus limiting the oxygen

consumption rate and the cooling demand of a culture. Commonly, by-product formation

(acetate) is suppressed by the limited substrate uptake rate in the fed-batch cultures. The

resulting cell concentration is usually higher than in a batch culture. Growth may be either

exponential or assume other rate profiles, depending on the flow rate. In this investigation, the

INTRODUCTION

2

fed-batch technique is further extended to include also non-limiting substrate feeds. The

continuous cultivation technique, (fig 1, right), has inlet streams for complete medium,

titrating agents and oxygen and an outlet stream for culture broth. In a continuous culture at

steady state conditions, the concentrations of all nutrients and the culture volume are constant

(Herbert et al. 1956). The specific growth rate is also constant at steady-state conditions and is

equal to the dilution rate if no cell death or recirculation occurs.

Figure 1. The three main techniques used for cultivation of microorganisms; batch (left), fed-batch (center),

and continuous cultivation (right). The inlet streams (arrows pointing in to the reactor) are for acids or bases

(H+/OH-), oxygen (O2) and feeds of substrates (F). The outlet stream (pointing out from the reactor) is for

culture broth (F).

One of the challenges for biotechnology processes is the limiting analytical techniques

available for evaluating product quality on-line or at-line (Larsson and Lam 2004). Today, the

ambition is to obtain the information from process data, rather than by end-point analysis, and

the degree of process understanding is reflected by the ability to predict the product quality.

The pharmaceutical industry is encouraged by the American Food and Drug Administration in

the process analytical technology framework to ensure high product quality by gaining

knowledge of the process and to identify its critical parameters (U.S Department of Health

and Human Services et al. 2004). Modeling as a tool for identifying such parameters, is of

special interest because of the low cost and time demand.

H+/OH-

O2 F F

F

H+/OH-

O2 H+/OH-

O2


3

Objectives

In this thesis, the temperature-limited fed-batch technique (TLFB) was developed to suppress

the excessive release of high concentrations of endotoxins found in the medium in

Escherichia coli fed-batch cultures at low specific growth rates and severe glucose limitation.

The productivity and physiology of the culture were evaluated in different modes of

cultivation.

Cultures producing a periplasmic protein and either grown with the TLFB or the glucose-

limited fed-batch (GLFB) technique were subjected to osmotic shock extraction. After the

treatment, the productivity was determined and the cytoplasmic membrane stability was

studied by analyzing DNA and soluble proteins in the extract. Furthermore, the solubility of a

recombinant, partially aggregated cytoplasmic protein was measured after production in

GLFB and TLFB cultures. A simulation model including the temperature-dependence of the

metabolic rates of glucose and oxygen uptake was developed. The parameters used in the

model were determined in pH-auxostat cultures at non-limiting substrate concentrations. The

model was further used for sensitivity analysis in order to identify critical parameters in the

culture.

INTRODUCTION

4


5

GROWTH CONTROL TECHNIQUES

In the present investigation, a fed-batch technique limited by the temperature was developed

for growth control (paper I-IV). The TLFB technique was compared to the commonly

applied GLFB technique (paper I, II). Furthermore, the pH-auxostat, a continuous cultivation

process, was utilized in paper III with temperature-limited growth control for the

determination of the temperature-dependent metabolic rates. The principles for the above

mentioned growth control techniques are addressed in this section.

To define how the concentration of a state variable (e.g. B) is influenced by process

conditions in a bioreactor, a mass balance for component B is usually written:

acc in out prod

!

dB

dt=Fin

V(B

i" B) +

Fout

V(B "#B) + r

B [g l-1 h-1]

(1)

Eq. 1 is fractionated as indicated in an accumulation (acc) term, an in- (in) and out-going

(out) term and a production term (prod). F is the flow rate either in or out from the fermenter,

V is the cultivation volume in the fermenter. δ is the fraction of B not recirculated, (δ=1 when

there is no recirculation). The volumetric reaction rate, rB, will here be rewritten as qB X. qB is

the specific production rate of B and X is the biomass concentration.

The mass balance (eq. 1) may be applied for all three modes of cultivation. Solving for the

state variables of biomass concentration (X), acetate concentration (A), glucose concentration

(S), dissolved oxygen tension (DOT), product concentration (P) and the volume (V) in a fed-

batch process and a continuous cultivation without recirculation of the component (δ=1) or

cell death may be written according to equations 2-7 (table 1).


6

Table 1. The mass balance equations of biomass (X), acetate (A), glucose (S), dissolved oxygen (DOT),

product (P), and volume (V) in a fed-batch process and in a continuous cultivation process with no cell

recirculation and no cell death.

Biomass

!

dX

dt=F

V("X) + µX (2)

Acetate

!

dA

dt=F

V("A) + qA X (3)

Substrate

!

dS

dt=F

V(Si " S) " qSX (4)

Dissolved oxygen

!

dDOT

dt= KLa(DOT *"DOT ) " qOXH (5)

Product

!

dP

dt= µ("P) + qP X (6)

Volume

!

dV

dt= F (7)

µ is the conventional denotification of the specific growth rate. qA, qS, qO, and qP are the

specific rates for acetate, glucose, oxygen and product, respectively. Eq. 6 describes the mass

balance for growth-associated products, e.g. intracellular products.

Eq. 5 describes the oxygen transfer from the air into the medium solution and the unit is [%

air sat. l-1 h-1]. Oxygen exhibits a limited diffusivity and solubility in aqueous solutions. The

volumetric oxygen transfer coefficient of the reactor (KLa) and is the dissolved oxygen

tension at equilibrium with the oxygen concentration in the gas bubble (DOT*) describe the

oxygen transfer capacity of the reactor. Henry’s constant (H) relates the dissolved oxygen

concentration in g l-1 to % air saturated. Parameters that influence the oxygen transfer capacity

are fermenter design (stirrer, baffles, sparger and vessel dimension), medium (oxygen

solubility, viscosity and surface tension), and the morphology of the microorganism.

For a facultative microorganism such as E. coli, oxygen limitation will cause the culture to

switch into mixed acid fermentation (Böck and Sawers 1996). The ATP generation is

significantly lower in anaerobic cultures and as a result, the rate of glycolysis increases to

obtain enough energy for growth. Large amounts of end products such as formate, succinate,

lactate, acetate and ethanol, which are not desired in the aerobic process, are formed. Thus,

various methods to control growth in high-cell-density cultures so that the oxygen

consumption rate of the culture does not exceed the oxygen transfer capacity were developed

(Lee 1996; Shiloach and Fass 2005).


7

Glucose limitation

Growth rate control by an essential substrate is performed with continuous feeding. The

feeding rate may be either preset or feedback controlled. The preset feeding profile is

typically preceded by a batch phase followed by either an exponential or a constant feeding

rate profile (Riesenberg et al. 1991).

Many different carbon and energy sources have been used as the growth-limiting substrate for

different production purposes in different microorganisms (Yamané and Shimizu 1984) but

glucose and other sugars are the most frequently utilized in fed-batch and continuous

cultivations. Glucose limitation is easy to control and below specific growth rates of 0.3 h-1,

the acetic acid from the overflow metabolism of E. coli is not produced (El-Mansi 2004;

Meyer et al. 1984).

In a GLFB culture with constant feed, the glucose concentration approaches quasi steady-state

(dS/dt≈0) and S << Si. Thus, the feeding rate of ingoing glucose must be lower than the

maximum glucose uptake rate of the cells:

!

FSi < qS,maxXV (8)

Figure 2A, shows a simulation of a GLFB process in which the glucose feed was feed-back

regulated by the DOT signal.

Temperature limitation (I)

In the TLFB technique, the glucose limitation in the GLFB culture is replaced by temperature

limitation (fig 2B). The temperature was regulated to keep the DOT at a set-point value (30

%) in order to limit the oxygen consumption rate of the culture. A gradually declining

temperature (T) controlled the growth rate (µ) and the oxygen consumption rate while all

other nutrients, including glucose, were in excess. It is a patented technique for the control of

endotoxin release (Silfversparre et al. 2002).

The glucose feed in a TLFB culture was controlled so that the glucose neither accumulated

too much (preferably dS/dt ≈ 0), still preventing excessive acetate formation, nor became

limiting, resulting in a GLFB process with endotoxin release. In the TLFB cultures an on-line

glucose analyzer monitored the glucose concentration and the feed rate was manually adjusted

(paper I-IV).


8

The glucose uptake rate was near maximum and the feeding rate was adjusted close to the

maximum glucose uptake rate of all the cells:

!

F (Si " S) # qS,maxXV (9)

Another feeding technique developed by Åkesson et al. (Åkesson et al. 2001) was applied for

TLFB cultures at other laboratories (de Maré et al. 2005; Ramchuran 2005). In this technique

the feed rate was pulsed up- or downward for short periods of time. The feeding rate was

regulated so that the monitored DOT signal did not respond to up-pulses but increased in

response to down-pulses. The interpretation of the response was that the glucose

concentration was high enough to obtain maximum respiration and low enough to avoid

acetate formation.

Temperature limitation has been applied for growth control also in continuous cultivations,

e.g. in the pH-auxostat (paper III). Bauer and White used temperature variations in batch

cultures and obtained higher cell concentration than in batches grown at a constant

temperature (Bauer and White 1976).

A: GLFB B: TLFB

0

10

20

30

40

50

0

20

40

60

80

100

0 5 10 15 20

X (

g l

-1),

F (

ml h

-1)

100 µ

(h-1), D

OT

(%)

time (h)

F

X

DOT

µ

0

10

20

30

40

50

0

20

40

60

80

100

0 5 10 15 20

X (

g l

-1),

T (

°C)

100 µ

(h-1), D

OT

(%)

time (h)

T

µ

DOT

X

Figure 2. Simulations of A: a GLFB and B: a TLFB culture. The glucose (A) and temperature (B) limitation

started at approximately 5 hours. The initial volume in the batch phase was 5 l. For the simulation model

description and parameters, see paper III. Feed rate (F), Biomass concentration (X), Dissolved oxygen tension

(DOT), Specific growth rate (µ), Temperature (T).


9

pH-auxostat (III)

The pH-auxostat is a continuous cultivation technique, only stable during growth conditions

when the concentrations of nutrients in the medium are non-limiting (Martin and Hempfling

1976). The technique has been applied for the determination of temperature-dependent

metabolic rates in different microorganisms (Larsson et al. 1990; Pham et al. 1999;

Satroutdinov et al. 2005; Yu et al. 2005), for metabolic studies (Zhao and Lin 2002), in the

dairy industry (Laht et al. 2002; MacBean et al. 1979), for production of filamentous fungi

(Simpson et al. 1995), for shear rate studies (Verschuren and Van der Heuvel 2002) and for

industrial waste-water treatment (Brune et al. 1982).

When E. coli uses ammonia as nitrogen source, a proton is generated:

α C6H12O6 + β NH4+ + γ O2 → δ CaHbOcNd + ε CO2 + φ H2O + β H+

glucose biomass

If other acids or bases are involved in the metabolism, the formula above must be

complemented. In an E. coli pH auxostat acetic acid is produced as a result of the overflow

metabolism:

α C6H12O6 + β NH4+ + γ O2 → δ CaHbOcNd + ε CO2 + φ H2O + η H+ + ι Ac-

Mass balances for the state variables in a pH-auxostat may be written according to eq. 1-7. A

simulation of a pH-auxostat run at different temperatures, resulting in different dilution rates,

is shown in figure 3A. In comparison, the more frequently used continuous cultivation

technique, the chemostat, controlled by different substrate feed rates (eq. 8) resulting in

different dilution rates is shown in figure 3B.

A pH-auxostat may be operated by either of two methods, (Siano 1999), by supplying the

glucose and the pH-control reagent with two separate inlet streams or by adding the pH-

control reagent into the fresh medium and supply this mixture in one single inlet stream. In

the present investigation, the latter of the two methods was applied. The driving force for the

inlet medium feed is the pH-difference of the inlet medium and the fermenter medium.


10

A: pH-auxostat B: Chemostat

0

2

4

6

8

10

0

20

40

60

80

100

0 20 40 60 80 100

X (

g l

-1)

10

0 µ

(h-1), D

OT

(%)

37 °C 21 °C25 °C29 °C33 °C

µ

DOT

X

time (h)

0

5

10

15

20

0

20

40

60

80

100

0 20 40 60 80 100

µ

DOT

X

time (h)

10

0 µ

(h-1), D

OT

(%)

X (

g l

-1)

0.6 h-1

0.5 h-1

0.4h-1

0.3 h-1

0.2 h-1

Figure 3. Simulations of A: a pH-auxostat, which is growth limited by different temperatures (37-

21 °C) and B: a chemostat, which is growth limited by glucose-limiting feed rate resulting in

different dilution rates (0.6- 0.2 h-1). For simulation model and parameters, see paper III. Biomass

concentration (X), Dissolved oxygen tension (DOT), Specific growth rate (µ).

A mass balance based on the proton concentration, H+ in the system may be written based on

equation 1:

!

dH+

dt=F

VNH

4

+"R

H+NH

4

+ # rH

+ (9)

!

NH4

+ is the ammonia concentration, used for nitrogen source and for neutralization. RH+/NH4+

is the stoichiometric conversion of proton from ammonia and rH+ is volumetric production

rate of protons. Because protons are generated when nitrogen is incorporated into biomass and

when acetate is produced, rH+ is the sum of the volumetric production rate of biomass rX and

acetate rA with respectively stoichiometric conversion factor, RH+/X and RH+/A:

!

rH

+ = rX RH +X

+ rARH +A

, and rX = µX and rA = qA X = µA

At steady state, there is no accumulation of protons, dH+/dt=0, and the specific growth rate is

the same as the dilution rate, µ = F / V. The concentration of ammonia at steady-state

conditions in a pH-auxostat can be calculated according to:

!

NH4+

= (RH

+XX + R

H+AA) "

1

RH

+NH4

+

(10)


11

Equation 10 implies that the added base concentration into the fresh medium feed controls the

acetate and biomass concentration at steady state. However, it is true only if all nutrients in

the fermenter are in excess. Inserting qS =µ / YX/S (YX/S is the yield of biomass per added

glucose), in the glucose mass balance (eq. 4) at steady-state conditions, the glucose

concentration in the inlet fresh medium is:

!

Sin

>X

YX S

(11)

If a lower inlet concentration of glucose was supplied to the medium, the stop-flow condition

described by (Larsson et al. 1990; Minkevich 1986; Siano 1999) would occur.


12


13

PHYSIOLOGICAL RESPONSES

A sudden increase of endotoxin release in GLFB cultures at low growth rates and severe

glucose limitation was observed. To investigate whether the release depended on the low

growth rate or the severe glucose limitation, the TLFB technique was utilized, which is

controlled by the oxygen consumption rate and the specific growth rate is limited by the

temperature instead of glucose. In the TLFB cultures, the endotoxin release was repressed

(paper I).

Some other physiological characteristics of TLFB cultures were observed in the present

investigation and will be discussed in this section. The shedding of endotoxins and the degree

of cell lysis increased significantly at a high agitation speed in a TLFB culture (paper IV).

Samples from TLFB and GLFB cultures were subjected to osmotic shock for extraction of a

periplasmic protein. Lower concentrations of soluble proteins and DNA were found in

extracts of cells grown at low temperatures than in extracts of cells grown at high

temperatures (paper II).

Stress responses

E. coli possesses several stress response systems assisting in the survival when the

environment becomes harsh (Enfors 2004). Several sensors, sensitive to different

environmental parameters, control the response genetically. The sensors trigger the

transcription of genes encoding for sets of proteins necessary for growth and survival under

the prevailing conditions. The environmental changes that E. coli is exposed to during culture

conditions are reflected in the resulting growth and productivity (Hoffmann and Rinas 2004);

therefore, the knowledge of how the cell responds physiologically to stress is important in

biotechnology.

The stringent response is induced by carbon and nitrogen (amino acids) limitations by

increasing the pool of guanosine tetraphosphate, ppGpp, which is an effector molecule that

adjusts the protein synthesis rate to the current conditions. The main target is to down-regulate

the growth-related gene transcription of tRNA and rRNA biosynthesis (Magnusson et al.

2005).


14

The general stress response is induced by e.g. reduced growth rate, high cell density, low

temperature, high osmolarity, low pH, carbon starvation and high temperature (Hengge-

Aronis 2002). The sensor, σS (RpoS), is able to induce a large number of genes relevant for

the protection of the cell.

Disturbances affecting the periplasmic space and the outer membrane (OM), e.g. antibiotics,

toxic compounds, heat or cold, induce the envelope stress response (Raivio 2005; Tam and

Missiakas 2005). The LPS and OM porins, which are involved in the maintenance, adaptation

and protection of the cell are synthesized upon induction. This response may also be

associated with pathogenic functions (Rhodius et al. 2006).

Cold shock

Upon cold shock, which is induced by a downshift in temperature, a transient arrest in growth

occurs due to a blockage at the translation initiation. To resume growth, a number of proteins

are induced, considered essential for the adaptation to the low temperature (Jones et al. 1987;

Kim et al. 2005). CspA, a major cold-shock protein, functions as an RNA chaperone, involved

in the re-stabilization of the nucleic acids after temperature-downshifts (Jiang et al. 1997).

Etchegaray et al. reported that levels of CspA were found after temperature shifts from 37

down to 30 °C, however the highest levels were found after shifts down to 20°C or 15 °C.

CspB, a homologue of CspA, was found only after shifts from 37 °C down to 20 °C or below

(Etchegaray et al. 1996). Thus, the thermoregulated cold-shock proteins CspA and CspB are

synthesized to the highest levels below 20 °C.

In fact, abrupt up- or down-shifts in temperature within the range 37-20 °C may be performed

without an arrest in growth rate (Ingraham and Marr 1996). Herendeen et al. performed

continuous cultivations of E. coli at temperatures ranging from 46-13.5 °C. The specific

growth rates of the continuous cultures formed a linear relationship in a log-scaled diagram

between 21-37 °C but deviated outside this interval (Herendeen et al. 1979).

When a culture is exposed to a lower temperature, it responds differently to a cold shock

(sudden down-shift) compared to a continuously declining temperature, to which a culture

may adapt smoothly. For instance, the culture may adapt by a decrease in growth rate and an

alteration in the membrane composition to maintain its fluidity, which is addressed further

below.


15

The E. coli cell and outer membranes

E. coli is Gram-negative, hence the cytoplasm is surrounded by two lipid-based membranes,

the cytoplasmic membrane (CM) and the outer membrane (OM), separated by the periplasmic

space containing the cell wall (fig 4). The two-layer phospholipid CM is responsible for the

regulation of uptake of nutrients and metabolites (Kadner 1996). The two-layer OM consists

of the inner and the outer leaflet, mainly containing phospholipids and lipopolysaccharides

(LPS), respectively. Due to the gel-like structure of the LPS molecule (Nikaido 2003), the

outer layer functions as a barrier against various harmful compounds such as digestive

enzymes (Nikaido 1996). The periplasm serves as a buffered space, separating the cell from

the local surroundings (Oliver 1996). It contains a number of enzymes along with components

destined for incorporation into the outer membrane (OM), or the cell wall. However the

protein concentration in the periplasm is much lower than that in the cytoplasm. Directing

recombinant proteins to the periplasm by secretion is therefore an advantage. The cell wall is

very rigid, able to withstand high osmotic pressure, extreme pH and temperature conditions

(Beveridge 1999). It consists of complex heteropolysaccharides and is covalently linked to

lipoproteins embedded in the OM (Park 1996).

Figure 4. A schematic figure of the cytoplasmic and the outer membranes surrounding the E. coli cell. The

periplasmic space contains the cell wall, which contributes to the robust structure of the cell.


16

The LPS consists of three parts: i) lipid A, the hydrophobic anchor, being the toxic part

associated with the immune response in mammals ii) the polysaccharide core, functioning as a

transport barrier because of their lipophilic and often negatively charged carbohydrate chains

and iii) the polysaccharide O-antigen (Raetz 1996). In pharmaceutical production, the removal

of LPS (the endotoxins) is of high priority since even very low traces cause immunogenic

responses in mammals when injected intravenously (Schletter et al. 1995). When inhalated,

endotoxins found in the aerosols cause pulmonary diseases (Johnston et al. 1998). The LPS

molecule is heterogenous and only partially soluble. It forms aggregates due to its amphiphilic

nature in both aqueous and organic solvents. Hence, this molecule is not easily removed and

the removal is especially difficult from biological products such as proteins, polysaccharides

and DNA. Examples of purification procedures involve ion-exchange chromatography, two-

phase extraction and affinity techniques (Petsch and Anspach 2000).

Temperature adaptation of the cytoplasmic and outer membranes

When a culture is exposed to lower temperatures, the fluidity of the cytoplasmic and outer

membranes of the cells decreases. Increased fluidity for cell adaptation is obtained by

increasing the proportions of unsaturated fatty-acids (UFA) at the expense of the saturated and

long-chain fatty-acids in both the phospholipid (Sinensky 1974) and the lipid A (Vorachek-

Warren et al. 2002) structures. UFAs possess a greater degree of flexibility, hence possessing

a lower phase transition temperature than saturated fatty acids. In table 2, the phase transition

temperatures of lipids derived from membranes of E. coli K-12 W3102, grown at three

different temperatures are listed (Sinensky 1974). The phase transition temperature for the

lipids decreases with decreasing growth temperature and this is related to the different

fractions of UFA in the lipids (Casadei et al. 2002; Cronan and Gelmann 1973; Sullivan et al.

1979).

Table 2. Phase transition temperatures from E. coli W3102 grown at different temperatures (Sinensky 1974).

Temperature of growth

Lipid phase transition temperature

43 °C 27 °C

30 °C 16 °C

15 °C -1 °C


17

Endotoxin release (I)

During normal growth, vesicles or fragments of OM components were released (Hoekstra et

al. 1976). The vesicles contained the same composition and content of phospholipids and

enzyme activity as the OM, however the content of lipoprotein molecules was less in the

vesicles. The lipoproteins in the OM are linked to murein in the cell wall, hence involved in

the envelope stabilization, and the authors concluded that the shedding of OM fragments does

not occur at the linkages. Wensink and Witholt found the same results and stipulated that the

OM vesicles are released when the OM is expanding faster than the underlying peptidoglycan

layer (Wensink and Witholt 1981). It was suggested that the shedding of OM components was

a result of the cell wall turnover in which peptidoglycan fragments from the cell wall were

excised (Beveridge 1999; Doyle et al. 1988; Zhou et al. 1998). Zhou et al. argued that if some

of the fragments were unable to penetrate the OM, they would probably be transported out

from the periplasm in vesicles. The OM forms blebs on the cell surface, which later are shed

off. The shedding of OM and cell wall fragments is mediated by autolysine, an enzyme which

was found in the vesicles analyzed by Zhou et al..

Certain factors may increase the rate of release of OM. Examples are phage infection (Loeb

and Kilner 1978), EDTA and heat treatment (Marvin et al. 1989; Pelltier et al. 1994; Tsuchido

et al. 1985), increased growth temperature (Mackowiak 1984), amino acid deprivation

(Ishiguro et al. 1986), antibiotics (Bucklin et al. 1994), citrate buffer and Tris buffer exposure

(Irvin et al. 1981).

High concentrations of endotoxins, 23 mg g-1 of biomass, were found in the medium in GLFB

cultures (paper I, fig 5 A). The extensive shedding corresponded to about 70 % of the total

lipopolysaccharide fraction of the E. coli cell (Neidhardt et al. 1990). Phospholipids and other

carbon containing outer membrane components were found in the medium at concentrations

corresponding to similar fractions as that of the endotoxins. The medium was analyzed for

soluble proteins and DNA, which could indicate cell lysis, however only low concentrations

were found. This was first observed by Han et al. by using a carbon mass balance (Han et al.

2003). The decreasing trend in the carbon recovery from initially 100% to about 90% based

on analyses of the biomass, carbon dioxide, acetate and glucose in three GLFB cultures

revealed that there were some carbon containing elements not included in the balance. The

concentrations of OM components determined in the medium corresponded to the missing

fraction.


18

A. GLFB culture 37°C B. TLFB culture

0

5

10

15

20

25

0

5

10

15

20

25

30

0 5 10 15 20

Endoto

xin

(m

g g

-1)

Bio

mass (g

l -1)

Time (h)

0

5

10

15

20

25

0

5

10

15

20

25

30

0 5 10 15 20

Time (h)

Endoto

xin

(m

g g

-1)

Bio

mass (g

l -1)

C. GLFB culture 20 °C

0

5

10

15

20

25

0

5

10

15

20

25

30

35

0 50 100 150 200

Endoto

xin

(m

g g

-1)

Bio

mass (g

l -1)

Time (h)

Figure 5. The endotoxin release (filled circles)

and biomass concentration (open circles) of

cultures grown with different techniques. A: the

GLFB technique at 37 °C. B: the TLFB

technique. C: the GLFB technique at 20 °C. Fed-

batch starts at about 5 hrs in A and B, and at 50

hrs in C. (Data compiled from paper I).

In TLFB cultures on the other hand, the endotoxin concentrations were very low, 1.3 mg g-1

of biomass (paper I, fig 5 B). The shedding of OM components in a GLFB culture increased

significantly at a specific growth rate of 0.1 h-1. The synthesis of LPS is under stringent

control (Ishiguro et al. 1986), which is induced at a late phase of GLFB cultures (Andersson

et al. 1996; Nyström 2004; Teich et al. 1999). However the endotoxins were not released as a

result of the low growth rate, as proven by the TLFB cultures.

To investigate whether the suppressed release of endotoxins in the TLFB cultures was caused

by the low temperature or by the high glucose availability, a GLFB culture was performed at

20 °C, which resulted in a high level of endotoxin, 20 mg g-1 of biomass (fig 5 C). Thus, the


19

conclusion was that the stability of the OM in TLFB cultures must be due to the glucose

availability, rather than the gradually decreasing temperature or the low growth rate (paper I).

The influence of stirring on endotoxin release (IV)

Ramchuran did not observe any difference between the endotoxin concentrations shedded into

the medium of an E. coli BL21 culture at the end of GLFB and TLFB processes (Ramchuran

2005). The culture conditions were approximately the same, however the agitation speed used

by Ramchuran was higher than what was used in this investigation.

Generally, it is assumed that a cell, such as E. coli is insensitive to the mechanical stress

exerted in bioreactors because their size is smaller in relation to the smallest eddies capable of

exerting turbulent flows (Thomas 1990). This microscale of turbulence is inversely

proportional to the energy dissipation rate, which is proportional to the stirrer rate, if the same

medium, reactor, stirrer and baffles are used (Chamsarta et al. 2005).

Still, there are some reports claiming that mechanical stress exerted from agitated bioreactors

effect bacteria and yeast. Examples of effects arising from higher mechanical stress reported

in the literature are an increased cell volume, (Edwards et al. 1989), a reduction of the general

transcription factor σB (Sahoo et al. 2003) and an inhibited growth and enzyme productivity

(Toma et al. 1991).

To investigate possible reasons for the higher endotoxin levels released in the TLFB cultures

of Ramchuran, two TLFB cultures were performed in this study (paper IV), one with higher

agitation speed, 1000 rpm, and one with lower agitation speed, 500 rpm, corresponding to

dissipation energies of 2 and 16,5 kW m-3, respectively. The resulting endotoxin concentration

at the higher stirrer speed was about 12 mg g-1 of biomass in the end (fig 6) compared to 23

mg g-1 of biomass in a GLFB culture (fig 5 A) and below 1 mg g-1 of biomass in a TLFB

culture with the lower stirrer speed (fig 5 B). The protein concentration in the medium was

also significantly higher in the culture with higher agitation speed (40 mg g-1 instead of 3 mg

g-1 at the low agitation speed) indicating cell lysis of about 10 % (paper IV).


20

0

5

10

15

20

25

0

5

10

15

20

25

30

35

40

0 5 10 15 20

Endoto

xin

(m

g g

-1)

Bio

mass (g

l -1)

Time (h)

Figure 6. The endotoxin release (filled circles)

and biomass concentration (open circles) of

cultures grown with the TLFB technique at 1000

rpm. Fed-batch starts at about 5 hrs. (Data

compiled from paper IV)

Hewitt et al. compared chemostat cultures of E. coli W3110 with different agitation speeds

(Hewitt et al. 1998). The viability or the cytoplasmic membrane potential of the cells for

different agitation speeds as measured with flow cytometry were the same. Analysis by

scanning electron microscopy indicated that an outer polysaccharide layer was stripped away

at the higher agitation speed. However, no effect was observed when treating the cells with

Rhodamin 123, a dye which stain cells with permeabilized OM. The conclusion drawn was

that the stability of the OM responded indifferently to the increasing agitation speed. In that

work, endotoxin analysis was not reported.

Temperature-dependent cytoplasmic membrane stability (II)

Constitutive production of periplasmic homologous β-lactamase was used as a reporter

protein in two GLFB cultures and three TLFB cultures (paper II). The periplasmic content

was extracted by the osmotic shock technique. It is a technique, which selectively extract

proteins of small native size, whereas large macromolecules are retained by the bacterial cell

(Vásquez-Laslop et al. 2001). The use of osmotic pressure to disrupt the OM in order to

retrieve periplasmic products without extensive lysis is a large-scale applicable technique

(Castan et al. 2002; Chaib et al. 1995). In the osmotic shock extraction, the cells are

suspended in a highly concentrated and cold (4°C) sucrose solution, creating an osmotic

pressure in the periplasm. It forces the cells to extract water from the cytoplasm into the

periplasm, which increases its volume so that the OM dissociates from the murein layer (Koch

1998). The hyperosmotic solution also contained EDTA, which permeabilizes the OM

(Pelltier et al. 1994), causing the OM to dissociate.


21

From the TLFB cultures, the average β-lactamase activity recovered from the extract of

samples withdrawn from the culture was about 9 kU g-1 of biomass (paper II). Soluble

proteins and DNA in the extracts were also analyzed. In figure 7, the β-lactamase purity, i.e.

the activity per mg of soluble protein, and the DNA concentration per g of biomass are shown

for one TLFB culture. The purity increased with increasing cultivation time and decreasing

temperature, from 100 to 450 U mg-1 protein. The DNA concentration in the extracts

decreased with increasing cultivation time and decreasing temperature.

In the two GLFB cultures the average β-lactamase activity recovered was 7 kU g-1 of biomass

(paper II). The purity of the target protein increased slightly with the cultivation time from

100-150 U mg-1 protein and the DNA concentration slightly increased from 2 to 3 mg g-1 of

biomass (fig 8).

The contaminations in the periplasm of soluble proteins and DNA, analyzed in samples

withdrawn from the TLFB cultures were lower than in extracts derived from the GLFB

cultures. Due to the DNA content, the contaminations probably originated from the cytoplasm

of cells with instable cell membranes. The endotoxin activity measured in the extracts was

similar in all cultures (paper II).

The results indicate that the cytoplasmic membrane (CM) of cells grown down to 18 °C was

more stable towards osmotic shock than cells grown at 37 °C. The membranes from cells

grown at 37 °C incorporate a higher amount of saturated fatty acids and when treating them at

lower temperatures, the membranes undergo a reversible phase transition, making them less

flexible and hence fragile (Sinensky 1974). Since the fraction of unsaturated fatty acids must

be higher in membranes from cells grown by the TLFB technique than from those grown by

the GLFB technique, it is possible that the lower accumulation of DNA and cytoplasmic

protein in the extract of cells grown by the TLFB technique is due to a more stable, i.e. more

fluid, CM.


22

0

10

20

30

40

50

0

2

4

6

8

10

0 6 12 18

T 6 feb

T 6 feb O_DNA (ug/mg dw)

DN

A (m

g g

-1 dw

)

time (h)

!-lacta

mase (

U m

g-1

pro

tein

), 1

0 *

tem

pera

ture

(°C

)

0

100

200

300

400

500

0

100

200

300

400

500

0

2

4

6

8

10

0 6 12 18

GLFB 09.15.44 06-09-19

O_specAkt (U/mg prot) O_DNA (ug/mg dw)

!-la

cta

ma

se

(U

mg-1

pro

tein

)

DN

A (m

g g

-1 bio

ma

ss)

Time (h) Fig. 7. The specific product activity (filled circles)

and the DNA concentration (open circles) in the

extract from osmotic shock treated samples, and the

cultivation temperature (crosses) from a TLFB

culture of E. coli accumulating β-lactamase in the

periplasm. Fed-batch starts at 0 hrs. (Data compiled

from paper II.)

Fig. 8. The specific product activity (filled circles) and

DNA concentration (open circles) in the extract from

osmotic shock treated samples, withdrawn from a

GLFB culture E. coli accumulating β-lactamase in the

periplasm. Fed-batch starts at 0 hrs. (Data compiled

from paper II.)


23

PROTEIN PRODUCTION

The TLFB technique may be an alternative to the GLFB technique for production of

recombinant proteins. The TLFB cultures exhibit similar cell productivity as the GLFB

cultures (paper I), which is an important parameter for intracellular production systems. A

higher productivity of the periplasmic protein β-lactamase was obtained in TLFB cultures

than in GLFB cultures (paper II). In addition, unpublished data (shown below) revealed that

the solubility of a partially aggregated recombinant protein was increasing initially after

induction in a TLFB culture.

The synthesis of recombinant proteins is usually controlled by a promotor, which is either

constitutive, allowing constant expression, or inducible, dividing the process into a cell

production and a protein synthesis stage. Promoters vary in strength and, if inducible, are

activated by different inducers. Generally, inducible promoters result in higher synthesis rates

of the protein. The production rate and the quality of the recombinant proteins depend on the

protein folding efficiency. Incorrectly folded proteins are either subjected to proteolysis,

hence lowering the net production rate, or will aggregate into inclusion bodies, hence forming

biologically inactive proteins (Murby et al. 1996). Aggregates must be refolded to regain

biological activity. However the recovery of inclusion bodies may be an attractive alternative

because inclusion bodies are easy to separate and the product concentration is high, usually

50-90 % (Fahnert et al. 2004).

Protein folding efficiency is enhanced by chaperones and foldases, however the steric

information necessary for the correct folding is in the primary structure of the nascent protein.

The macromolecular concentration (RNA and proteins) in the cytosol is very high (340 mg

ml-1), which enhances the misfolding of proteins compared to the folding in much more dilute

samples in a test tube (Ellis and Hartl 1999). Yet there are reports of enhanced efficiency of

chaperones and foldases due to the high concentration in the cytoplasm. The functionality of

the refolding catalyst protein disulfide isomerase, PDI, increased its efficiency during

crowded conditions (van den Berg et al. 1999).

PROTEIN PRODUCTION

24

Partial aggregation is an unwanted effect caused by insoluble and incorrectly folded proteins,

resulting in reduced biological activity. Heterologues recombinant proteins tend to form

aggregates to a larger extent, especially those from eukaryotes. The translation rate in

prokaryotes are 5-10 times faster than in eukaryotes and the folding is therefore performed

post-translationally in prokaryotes (Ellis and Hartl 1999). They are unable to glycosylate,

which is a feature in higher organisms that enhances the solubility. The oxidation of disulfide

bonds between thiol-groups of cysteines in prokaryotes is hampered by the lower redox

potential of the cytoplasm compared to that of the endoplasmic reticulum of eukaryotes. Thus,

the aggregates formed in the cytoplasm of prokaryotes might be mainly due to the absence of

correct disulphide bonds (Fahnert et al. 2004).

Several strategies of minimizing the aggregate formation and proteolytic activity of

heterologous proteins are presented in the literature. It is an advantage of secreting the product

into the less crowded compartment of the periplasm (van den Berg et al. 1999). Since the

synthesis occurs in the cytoplasm, the product must be tagged by a signal peptide sequence to

be able to be secreted to the periplasm. Here, the spontaneous formation of disulphide bonds

is enhanced due to the slightly oxidizing environment (Fahnert et al. 2004). In addition, the

proteins are less subjected to proteolysis in the periplasm, since most proteases are found in

the cytoplasmic compartment (Rozkov and Enfors 2004).

Low temperature is known to increase solubility of many recombinant proteins. Schein and

Noteborn showed that the solubility of proteins increased when they were produced at 28-30

°C instead of at 37 °C (Schein and Noteborn 1988). The reaction rate is a function of the

temperature as described by the Arrhenius relationship. Hence, protein synthesis, protein

folding and self-association is decreased at the low temperatures (Georgiou and Valax 1996).

The proteolysis rate is also a temperature-dependant parameter. The proteolysis rate for

degrading ZZT2, produced in the cytoplasm in E. coli, was significantly lower for cultures

grown with the TLFB technique, compared to cultures grown with the GLFB technique

(Rozkov 2001). The thermodynamics of the proteolysis reaction rate was supposed to be

responsible for the decrease. Very low specific glucose uptake rates (Boström et al. 2005; Lin

et al. 2004) and high concentrations of inducer (Choi et al. 2000; Ramirez and Bentley 1999;

Sandén 2004) are examples of other culture condition related variables with a resulting

negative effect on productivity. High acetate concentrations inhibit both growth (Han et al.

1992; Luli and Strohl 1990; Xu et al. 1999) and productivity (Konstantinov et al. 1990;

Turner et al. 1994).


25

In intra-cellular production systems, the specific growth rate is closely related to productivity

(eq. 6). The cell production rate was the same in the TLFB and the GLFB cultures and the

cell-density obtained was the same (paper I, II). The biomass concentrations of a GLFB and

a TLFB culture are compared in figure 9 and the techniques are equally efficient with respect

to cell productivity. In high cell-density fed-batch cultures applying glucose or temperature

limitation for controlling the oxygen consumption rate, the degree by which growth is limited

depends on the oxygen transfer coefficient of the fermenter, KLa. Indeed, the time required to

produce the same amount of cells is the same by both techniques.

0

5

10

15

20

25

30

0

5

10

15

20

25

30

0 5 10 15 20

Bio

mass (

g l

-1)

Endoto

xin

(mg g

-1 bio

mass)

time (h)

Figure 9. Biomass concentration

(squares) and the released endotoxin

concentration related to biomass (mg g-1

of biomass, circles) in a TLFB culture

(filled symbols) and a GLFB culture

(open symbols). Fed-batch start at

approximately 5 hrs in both cultivations.

(Result from paper I.)

The significant difference in endotoxin concentration, shown in fig 9 and addressed in the

previous section, would favor the TLFB technique for producing extra-cellular products due

to the resulting reduced purification costs. However for extra-cellular production, the cell

productivity is not always related to protein productivity. Very few examples of extracellular

production systems in E.coli exist for recombinant proteins. If the protein is to be transported

over the OM, E. coli exhibits natural secretion mechanisms to export toxic compounds and

some of those may be used for transport of recombinant proteins, but it may also occur by

leakage of periplasmic proteins (Mergulhao et al. 2005; Shokri et al. 2003).

Production rate and protein solubility (II)

The periplasmic protein β-lactamase was constitutively produced in two GLFB cultures and

three TLFB cultures (paper II) and was recovered by osmotic shock as described in the

PROTEIN PRODUCTION

26

previous section. In this section the production rate of β-lactamase is discussed. In figure 10,

the specific production rates of average concentrations of active β-lactamase in three TLFB

and two GLFB cultures are shown. The product activity was measured in disintegrated culture

samples, and the specific productivity was calculated according to eq. 6.

0

500

1000

1500

2000

2500

0 0,1 0,2 0,3 0,4

sp

ecific

pro

du

ctio

n r

ate

(U

g-1

h-1

)

specific growth rate (h-1

)

Figure 10. The specific production rate of β-

lactamase as a function of the specific growth

rate. Average values of three TLFB cultures

(filled circles) and two GLFB cultures (open

circles). (Data compiled from paper IV.)

The specific production rate was higher in the TLFB cultures than in the GLFB cultures. The

synthesis burden for the cell was very low, 2-3 mg of β-lactamase g-1 of biomass (conversion

factor 3550 U mg-1 of β-lactamase, (Georgiou et al. 1988), and the protein is homologous,

thus the risk of misfolding or protein degradation due to proteolysis should be very small.

Different plasmid content in the TLFB and the GLFB cultures is one possible reason. Another

reason could be that the synthesis rate was higher in the TLFB cultures. Similar observations

were made in the production of the fusion protein CBD-lipase in Pichia pastoris in methanol-

limited fed-batch and TLFB cultures (Jahic et al. 2003).

The fact that the temperature gradually declines in the TLFB process indicates that one would

expect an increasing solubility of a protein that is known to form inclusion bodies. In

cooporation with AstraZeneca, a recombinant protein that yielded about 70-80% inclusion

bodies (according to scanning of Western blots) was investigated. When the solubility was

compared in a GLFB and a TLFB process, it was much higher only during the early phase of

the TLFB process. Then, it gradually declined towards 40 % at the end of the culture (fig 11),

opposite to what is expected from the temperature effect on recombinant protein solubility


27

(Schein and Noteborn 1988). A factor that influences the interpretation of the results is

proteolysis, although not performed in this study, it would increase the ratio of inclusion

bodies. If the rate of proteolysis increased with process time, along with the concomitant

decrease in temperature, the decrease of the soluble fraction could possibly be explained by

temperature-dependent induction of new proteases (Rozkov and Enfors 2004).

0

20

40

60

80

100

0 5 10 15 20 25

So

lub

le f

ractio

n (

%)

time after induction (h)

Fig 11. The solubility fraction of the recombinant

protein produced in a TLFB (filled circles) and a

GLFB (open circles) after induction (unpublished

data).

PROTEIN PRODUCTION

28


29

MODELING THE TEMPERATURE-DEPENDENT RATES

Modeling as a tool for identifying critical parameters of a process, is of special interest

because of the low cost and time demand. The development of a kinetic model for E. coli

cultured by the TLFB technique includes the determination of the temperature-dependent

metabolic rates (paper III). Sensitivity analysis to gain knowledge of the critical parameters

and their impact on productivity was investigated by using the developed model (paper IV).

The oxidative and overflow pathways

Figure 12 depicts the glucose energy metabolism in E. coli grown aerobically on mineral

medium. Glucose is the most common carbon source for microorganisms in pharmaceutical

biotechnology. This molecule generates both energy and precursors for biosynthesis. Glucose

is transported into the E. coli cell by the phosphoenolpyruvate phosphotransferase system

(Postma et al. 1996). The Embden-Meyerhof pathway (EMP) and the pentose-phosphate

pathway (PPP) are the two trunk routes for the oxidation of glucose into pyruvate (Fraenkel

1996). The intermediates in the glycolysis are precursors for amino acids, fatty acids and

nucleic acids. In addition, 2 ATP are generated and 2 NAD+ are reduced to NADH from each

glucose molecule. Pyruvate is converted into acetyl-CoA during aerobic conditions either by

the oxidative or the so-called overflow metabolic pathway. The path selected depends on

substrate accessibility and other environmental conditions.

In the TCA cycle, acetyl-CoA is oxidized into carbon dioxide. The main oxidant, NAD+, is

reduced into NADH. The reducing agents from the TCA cycle and from the EMP are then

reoxidized in the respiratory pathway, where oxygen is reduced into water with concomitant

generation of ATP (fig 12). Totally, about 22 ATP are generated in the oxidative metabolism.

When the glucose flux to acetyl-CoA exceeds the capacity of the cell to further metabolize it

in the TCA cycle, the surplus acetyl-CoA is converted to acetate (fig 12). In this pathway, one

ATP is generated but no coenzymes are reoxidized. The overflow metabolism is observed in

batch (Luli and Strohl 1990), in fed-batch (Konstantinov et al. 1990; Luli and Strohl 1990),

and continuous cultivations (Han et al. 1992).


30

Acetate is also formed under oxygen limitation in the so-called mixed acid fermentation.

However, this reaction takes another path from pyruvate, which is split to formate and acetyl-

CoA from which the acetate is formed. Concomitantly a number of other products are formed:

ethanol, lactate, formate and succinate. None of these products is formed in the overflow

metabolism (Andersen and von Meyenburg 1980; Xu et al. 1999).

Figure 12. Summary of E. coli energy metabolism (not stoichiometric). The glucose is oxidized into acetyl-CoA,

while generating reducing agents, CO2 and ATP. If the specific glucose uptake rate is below a critical value, the

carbon from glucose is oxidized into carbon dioxide while generating reducing agents, e.g. NADH. NADH is

reoxidized into NAD+ in the respiration with concomitant ATP production. The eventual surplus of acetyl-CoA,

not oxidized in the TCA cylce is converted to acetate in the overflow metabolism with an energetic gain of

approximately 1 ATP. The total energetic gain of one molecule of glucose metabolized in the oxidative

metabolism is about 22 ATP.

The maximum uptake rate of oxygen for E. coli W3110 is about 13-14 mmol g-1 h-1 (Xu et al.

1999). Above a critical specific glucose uptake rate the oxygen consumption rate remains

constant, while a further increase in glucose concentration results in a higher glucose uptake

rate. It has been disputed whether the metabolic bottleneck of the overflow metabolism of E.

!

Glucose

!

Acetyl - CoA

!

Acetate

!

ADP

!

ATP

!

Pyruvate

!

TCA

!

CO2

!

NADH

!

NAD+

!

ADP

!

ATP

!

O2

!

H2O

!

ADP

!

ATP

!

NAD+

!

NADH

!

NAD+

!

NADH

!

Overflow metabolism

!

Oxidative metabolism

CO2


31

coli and S. cerevisiae lies in the TCA cycle or in the respiratory pathway (Andersen and von

Meyenburg 1980; Sonnleitner and Käppeli 1986; Verduyn et al. 1992). Irrespectively of

where the bottleneck is located, a maximum in the specific oxygen uptake rate, qOmax, is

observed both in S. cereviciae (Pham et al. 1999) and in E.coli (Xu et al. 1999) when the

glucose uptake rate exceeds a critical value.

Mathematical model (III)

The feed of glucose was set to approximately support maximum growth rate at the process

temperature in the pH-auxostats and the TLFB cultures. Thus, the in-going glucose flux was

divided into both the oxidative and the overflow metabolic pathway (fig 12). Xu et al.

presented a model of the metabolism of E. coli including both production and consumption of

acetate (Xu et al. 1999). Consumption of acetate occurs when the culture is switched from

batch mode into a limiting glucose supply. This model (without acetate consumption) was

used in paper III and is presented in fig 13.

Figure 13. Model describing the in-going substrate fluxes of the glucose, S, and oxygen, O2, and the

produced out-going fluxes of carbon dioxide, CO2, biomass, X, and acetate, A. The model equations

describing the rates by which the glucose or the oxygen is going through either the oxidative or the overflow

metabolism and the nomenclature used are described in the text (from paper III).

!

O2

!

qO = qS,ox,enRO / S " qO,max

!

qS = qS,maxS

S + KS

!

qS,of ,an = qS,ofYX / S,ofCX

CS

!

qS,ox

!

qS,of = qS " qS,ox

(if qO = qO,max)

!

qS,ox,en = qS,ox " qS,ox,an

!

qS,of ,en = qS,of " qS,of ,an

!

S

!

qA = qof ,enRA / S

!

qS,ox,an = qS,oxYX S,ox

CX

CS

!

µ = qS,oxYX S,ox + qS,ofYX / S,of

!

CO2

!

X

!

A


32

The glucose uptake rate is assumed to follow Monod kinetics:

!

qS = qS,maxS

S + KS

(12)

qS is the specific rate by which the culture takes up glucose, qSmax is the maximum qS. S is the

concentration of glucose in the medium and Ks is the saturation constant for glucose uptake.

At low glucose concentrations all glucose is metabolized by the oxidative metabolism.

!

qS,ox = qS (13)

The flux to anabolism (qS,ox,an, eq. 14) is obtained from a carbon mass balance. The rate of

carbon uptake to the anabolism, qS,ox,anCS (g carbon per g biomass per hour) equals the rate of

carbon accumulation in the cells, qS,oxYX/S,oxCX. CS and CX are the carbon concentrations in

glucose and biomass, respectively.

!

qS,ox,an = qS,oxYX S,ox

CX

CS

(14)

The glucose flux for the energy metabolism becomes:

!

qS,ox,en = qS " qS,ox,an (15)

The respiration rate for re-oxidation of the co-enzymes (Fig 12) is dependent on the glucose

flux for energy metabolism, which is combusted into carbon dioxide:

!

qO = qS,ox,enRO / S (16)

The coefficient RO/S is the stoichiometric ratio between oxygen and energy substrate in the

respiration (6 moles of oxygen per mole of glucose).

The glucose flux towards the overflow metabolism can be calculated with the concept of

maximum respiration rate (qO,max). Thus, if the rate of respiration (eq. 16) becomes higher

than qO,max, the fluxes to oxidative anabolism (qS,ox,an) and oxidative energy metabolism

(qS,ox,en) are adjusted to fit qO,max and the remaining total substrate uptake is directed to

overflow metabolism:

!

qS,of = qS " (qS,ox,an + qS,ox,en ) (17)


33

The acetylCoA metabolized from glucose (fig 12) destined for the overflow metabolism (eq.

17) is converted to acetate with concomitant generation of ATP, providing energy for

additional growth, although with a lower yield coefficient (YX/S,of) than that of the oxidative

growth (YX/S,ox). The specific growth rate becomes:

!

µ = qS,oxYX S,ox + qS,ofYX / S,of (18)

The glucose fluxes into anabolism (eq 19) and energy metabolism (eq. 20) in the overflow

metabolism is further modeled in analogy with the fluxes of the oxidative metabolism:

!

qS,of ,an = qS,ofYX / S,ofCX

CS

(19)

!

qS,of ,en = qS,of " qS,of ,an (20)

The rate of acetate formation, qA is obtained from the stoichiometric ratio of glucose

conversion to acetate (RA/S) of the flux diverted to energy metabolism:

!

qA = qS,of ,enRA / S (21)

The rate equations 12, 16, 18 and 21 were inserted in corresponding mass balance equations

(table 3) for the dynamic simulations of both the pH-auxostat and the TLFB process.

Table 3. The equations used for the dynamic simulations of both the pH-auxostat and the TLFB process.

state variable mass balance rate equation

biomass

!

dX

dt= X ("

F

V+ µ)

!

µ = qS,oxYX S,ox + qS,ofYX / S,of

acetate

!

dA

dt= A("

F

V) + qA X

!

qA = qS,of ,enRA / S

glucose

!

dS

dt= (Si " S)

F

V" (qS,ox + qS,of )X

!

qS = qS,maxS

S + KS

dissolved oxygen

!

dDOT

dt= KLa(DOT *"DOT ) "HqOX

!

qO = qS,ox,enRO / S

Temperature-dependent metabolic rates (III)

To use the model equations for pH-auxostat or TLFB process simulations (fig 13 and table 3),

the temperature-dependent maximum specific rates of oxygen and glucose were determined in

pH-auxostat cultures (paper III).


34

The specific growth rate (µ) describes a linear relationship with the reciprocal of the

temperature in a log-scaled diagram within the temperature range 20-37 °C (Herendeen et al.

1979). It is a relationship known as the Arrhenius equation (eq. 22). k is a constant, specific to

each reaction, Ea is the energy of activation, G is the general gas constant and T is the absolute

temperature.

!

µ = keEa GT (22)

The specific rates were experimentally determined in a pH-auxostat (paper III) and were

plotted in an Arrhenius plot (eq. 22). The specific uptake rates of glucose and oxygen (Fig

14A) and the specific production rates of growth, acetate and carbon dioxide (Fig 14B)

deviated from the linear relationship described by Arrhenius at 18 °C.

A: The specific rate of glucose and oxygen B: The specific rate of biomass, acetate and CO2

0,01

0,1

1

10

3,23,253,33,353,43,45

Specific

rate

(g g

-1 h

-1)

Reciprocal temperature 1000 T-1 (K-1)

18 °C

20 °C

23 °C

26 °C

28 °C

33 °C

37 °C

qS

qO

0,001

0,01

0,1

1

0,01

0,1

1

10

3,23,253,33,353,43,45

Specific

rate

of C

O2 (g

g-1

h-1)

Spec. ra

te o

f A

(g g

-1 h

-1)

and X

(h

-1)

Reciprocal temperature 1000 T-1 (K-1)

18 °C

20 °C

23 °C

26 °C28 °C

33 °C37 °C

µ

qA

qCO2

Figure 14. Arrhenius plots of the maximum specific rates as a function of the reciprocal temperature. A:

glucose, qS (open triangles) and oxygen, qO (filled triangles). B: growth, µ (open circles), carbon dioxide,

qCO2 (open triangles) and acetate, qA (filled squares). Solid lines represent exponential fits according to the

Arrhenius equation in the range 20-37 °C. Dashed lines connect the fitted lines with the corresponding points

at 18 °C. The deviation from the fitted line was largest at 18 °C (from paper III).

Since the TLFB process runs with a temperature limitation below 20 °C for a significant

period of time (fig 15), a polynomial fit of the 3rd order of the experimental data was used

instead of the Arrhenius equation to describe the temperature-dependent maximum specific

rate of glucose uptake. The maximum specific rate of oxygen uptake was described as a linear


35

function of that of glucose (fig 4 in paper III). A simulation and corresponding experimental

data are shown for a TLFB culture in figure 15.

In the culture performed, the DOT was regulated around 30 % air. sat. (data not shown) by the

temperature. The temperature profile decreased from 37 °C down to 18 °C, the biomass

reached 27 g l-1 and the specific growth rate gradually declined to 0.02 h-1 (Fig 15A). The

glucose was manually regulated to a concentration of about 0.2 g l-1, the final acetate

concentration was about 1 g l-1 and the feeding rate was about 40 ml h-1 (Fig 15B). The initial

volume was 5 l.

The conclusion is that the simulations (Fig 15) fit reasonably well with experimental data. It

may be used to predict the temperature profile (Fig 15A) and a suitable glucose feeding rate in

order to avoid acetate accumulation (fig 15B). The temperature profile together with the

oxygen consumption rate is needed for estimation of the cooling demand (paper III).

A: temperature biomass and the specific growth rate B: Glucose and acetate concentration and feeding rate

0

10

20

30

40

0

0,2

0,4

0,6

0,8

1

0 4 8 12 16

Te

mp

era

ture

(°C

) a

nd

bio

ma

ss (

g l

-1)

Sp

ecific

gro

wth

rate

(h-1)

time (h)

X

µ

T

0

20

40

60

0

1

2

3

0 4 8 12 16

Fe

ed

ing

ra

te (

ml h

-1)

Ace

tate

an

d g

luco

se

(g l -1)

time (h)

F

A

S

Figure 15. A simulation of a TLFB culture and experimental data of the The temperature decreased from 37

°C down to 18 °C. A: Experimental data of the biomass (filled circles, X) growth rate (open circles, µ) and

temperature (crosses, T). B: Experimental data of the feeding rate (dashed line, F), glucose concentration

(dotted line, S) and acetate (squares, A). The volume was 5 l at fed-batch start (2 h). The simulated

parameters are indicated as solid lines. (Data retrieved from paper III.)

Figure 16 illustrates that the glucose uptake rate influenced the acetate formation differently

in cultures performed by the three different techniques, the pH-auxostat, the TLFB, and the

GLFB technique. In the pH auxostat there was a large excess of glucose in the medium. Thus,


36

the specific glucose and oxygen uptake rates were maximum, (qS,max and qO,max). About 10 %

of the carbon taken up by the cell was incorporated in acetate (paper III), hence the rest was

going to the oxidative metabolism. The TLFB culture also exhibited an excess of glucose

allowing qO = qO,max, however the excess glucose concentration in the medium was low to

limit the acetic acid accumulation resulting in a low flux for the overflow metabolism and in a

lower acetate production (about 2 % of the total carbon). The GLFB culture was limited by

the glucose supply, hence the respiration did not reach its maximum capacity.

The acetate production may also be avoided by choosing a low-producing strain. For instance,

E. coli B strains, frequently used by the industry, are low acetate producers and hence less

sensitive to glucose accumulation, while the E. coli K12 strains, used in the present

investigation, and also used by the industry, are high acetate producers (Phue and Shiloach

2004).

Figure 16. Schematic models over glucose fluxes in the cell to the different metabolic pathway destinations in

three different cultivation techniques: pH auxostat, TLFB and GLFB.

Sensitivity analysis performed by simulations (IV)

When two different agitation speeds in TLFB cultures were applied, different temperature

profiles and cell concentrations were obtained (paper IV). Sensitivity analysis was performed

to investigate possible reasons behind the differences. Variables investigated were process

dependent, the oxygen transfer coefficient KLa and the glucose concentration (S), or related to

the cell physiology, the specific cell death rate kd and the maximum specific glucose uptake

rate, qSmax (fig 17).

!

pH auxostat

qS " qSmax

qO = qO,max

!

oxidative metabolism

qSox = qSox,an + qSox,en

!

overflow metabolism

qSof = qS " qSox

!

acetate production

qA max

!

TLFB

qS =qSmax

(S + Ks)

qO = qO,max

!



!

overflow metabolism

qSof = qS " qSox

!

acetate production

qA

!

GLFB

qS =qSmax "S

(S + Ks)

qO < qO,max

!



!

no overflow metabolism


37

The KLa is dependent on the stirrer rate, the aeration rate and the liquid volume in the reactor.

Hence it is a variable that is controlled by the operator. In paper I it was shown how the final

biomass concentrations and the temperature profile in TLFB cultures depend on the KLa.

Figure 17 A shows two over-lapping simulations, with KLa of 270 h-1 and 470 h-1, of the

increasing final biomass concentrations and the temperature profiles. The temperature profile

remained the same and the biomass productivity increased with increasing KLa.

The excess glucose concentration, S, may be varied in the process. In the simulations, the

temperature profile was somewhat sharper with higher S (1.0 g l-1 instead of 0.2 gl-1, fig 17B).

However, the difference in temperature is insignificant for small differences in S, (e.g. 0.1 gl-

1, paper IV). More importantly, the acetate accumulation is higher for higher S, as explained

above (fig 16) and indicated in paper III. The biomass productivity, assuming not inhibited

by acetate, was not influenced by the glucose concentration.

The specific death rate of the first order, kd, was inserted in the mass balance for biomass (eq.

2) according to equation 23. If given the magnitude of 0.02 h-1, the biomass concentration was

decreased by about 10 % and the temperature profile was less steep, with about 1 °C

difference at 15 hours after TLFB start (Fig 17C).

!

dX

dt= "

F

VX + µX " k

dX (23)

qS,max is another organism-dependent parameter. If a non-competitive inhibitor is present, such

as acetic acid, qS,max may also be reduced. If the temperature-dependent function of qS,max,

derived in paper III, is reduced to half, the temperature profile becomes less steep (fig 17D).

The biomass productivity was similar, however the batch phase was shorter with higher qSmax.

Thus, the biomass curve and the specific growth rate are mainly determined by the KLa, which

does not influence the temperature profile. Parameters influencing the specific uptake rate of

glucose; S and qSmax, influences the temperature but not the biomass concentration profile. The

specific cell death rate, kd influences both the biomass concentration and the temperature

profile.


38

Figure 17. Analysis of the effect on the temperature (T) profile and the biomass concentration (X) from

changes in the culture condition related parameters A: the oxygen transfer coefficient (KLa) and B: the

glucose concentration (S); and changes of the physiological parameters C: the specific cell death rate (kd),

and D: the maximum specific uptake rate (qsmax). The values of the parameters are indicated above the graphs

respectively. qsmax= f(T)= 1.44- 0.186T +0.00754T2-0.0000693T3 (paper III). (From paper IV)

The sensitivity analysis in paper IV resulted in an increasing understanding of the TLFB

process, demonstrating the usefulness of a simulation model. It indicates that the temperature

profile may be used as a control parameter. If the temperature profile deviates from the

normal slope (empirically determined), the possibility of an altered culture physiology

prevails. Such physiological conditions could be infection, plasmid loss, inhibitors or cell

A. KLa= 270 h-1(thick lines) and 470 h-1 (thin lines) B. S= 0.2 g l-1(thick lines) and 1 g l-1 (thin lines)

0

10

20

30

40

50

60

20

25

30

35

40

-5 0 5 10 15

Bio

ma

ss (

g l

-1)

Te

mp

era

ture

(°C)

time (h)

KLa

T X

0

5

10

15

20

25

30

35

40

20

25

30

35

40

-5 0 5 10 15

Sset= 1G/L.txt

Bio

mass (

g l

-1)

Tem

pera

ture

(°C)

time (h)

S

X

T

C. kd= 0 (thick lines) and 0.02 h-1 (thin lines) D. qsmax= f(T)(thick lines) and 0.5*f(T) (thin lines)

0

10

20

30

40

50

20

25

30

35

40

-5 0 5 10 15 20

TLFB III.txtT

em

pe

ratu

re (°C

)Bio

ma

ss (

g l

-1)

time (h)

T

X

Kd

0

10

20

30

40

50

60

20

25

30

35

40

-5 0 5 10 15 20 25 30 35

Bio

ma

ss (

g l

-1)

Te

mp

era

ture

(°C)

time (h)

qSmax

T X


39

death. The concentrations of endotoxins and proteins corresponded to cell lysis of at least 10

% of the total cell concentration in the culture with the unusual temperature profile (paper

IV). The sensitivity analysis implies that cell death of this magnitude may contribute to the

different temperature profiles obtained.


40


41

CONCLUDING REMARKS

The TLFB cultivation technique offers several advantages over the more frequently used

GLFB technique for production of recombinant proteins in small to medium scale processes.

The main feature of the technique is the repressed release of the high amounts of endotoxins.

Consequently, the health risks associated with the exposure to aerosols containing endotoxins

from the outlet air of the bioreactor should be reduced. Furthermore, the purification costs of

extracellular products should be reduced, although few such recombinant proteins are

produced using E. coli as the host. However many recombinant proteins are secreted to the

periplasmic space. Thus, the increased stability of the cytoplasmic membrane obtained in the

cells grown with the TLFB technique, offers the benefit of less contaminated osmotic shock

extracts, with respect to DNA and protein concentrations. Yet, the endotoxin concentration

was equally high in extracts from GLFB and TLFB cultures.

A higher solubility fraction was obtained when the TLFB technique was applied to a

production process forming partially inclusion bodies. Unpredictably, the solubility soon

decreased to the same level as the reference, which reflects the necessity of further

investigations in this matter. Many processes use a one-step reduction of the temperature to

increase the solubility of partially aggregated recombinant proteins. The TLFB technique

offers a smoothly declining temperature and a specific glucose uptake rate near maximum,

reassuring high production of energy for protein synthesis. Furthermore, although not studied

in this project, the degradation rate of proteolytically susceptible proteins should be lower for

thermodynamic reasons, which could be an interesting task for prospect investigations.

The apparent drawback for the TLFB technique is the increasing cooling demand, since the

temperature gradient between the cooling water and the bioreactor gradually declines. The

main heat evolution in an aerobic bioreactor is caused by the respiration. The temperature

profile and the respiration rate can be calculated from the mathematical model developed in

this study. Thus, the TLFB model offers a means for the calculation of the cooling liquid

demand, which adds to the production cost.

CONCLUDING REMARKS

42

The significantly increased endotoxin release in a TLFB culture at a higher agitation speed

was unexpected due to the small size of the E.coli cell. However, such energy dissipation

rates (16 kW m-3) are never reached in a large-scale bioreactor. Thus, this phenomena is not

necessary a limitation, although it should be further investigated before considering scale-up.

The results of the present investigation also demonstrate the feasibility of using a simulation

tool for sensitivity analyses for the purpose of gaining insight into which parameters, both

process control variables and physiological parameters, that may influence the performance of

a cultivation process.


43

NOMENCLATURE

Symbol Unit Parameter

A g l-1 acetate concentration

B g l-1 Concentration of an arbitrary state variable B.

CS g g-1 carbon fraction in glucose

CX g g-1 carbon fraction in biomass

DOT % air sat. dissolved oxygen tension

DOT* % air sat. DOT in equilibrium with the oxygen in the gas bubble

Ea J mol-1 energy of activation

F l h-1 flow rate

G J K-1 mol-1

Avogadro’s gas constant

H % air sat. l g-1

Henry’s constant

H+ g l-1 proton concentration

k h-1 reaction specific constant in Arrhenius (eq. 22)

KLa h-1 volumetric oxygen transfer coefficient

KS g l-1 saturation constant for glucose

qA g g-1 h-1 specific acetate production rate

qO g g-1 h-1 specific oxygen consumption rate

qP g g-1 h-1 specific product production rate

qS g g-1 h-1 specific glucose uptake rate

rA g l-1 h-1 volumetric reaction rate of acetate

RA/S g g-1 stoichiometric ratio of acetate per glucose in overflow flux

rB g l-1 h-1 volumetric reaction rate of an arbitrary state variable B.

Symbol Unit Parameter

rH+ g l-1 h-1 volumetric reaction rate of protons.

RH+/A g g-1 stoichiometric ratio of acetate per proton

RH+/NH4+ g g-1 stoichiometric ratio of ammonia per proton

RH+/X g g-1 stoichiometric ratio of biomass per proton

RO/S g g-1 stoichiometric ratio of oxygen consumed per glucose in respiration

rX g l-1 h-1 volumetric reaction rate of biomass

rpm h-1 stirrer speed

S g l-1 glucose concentration in bioreactor

Si g l-1 inlet glucose concentration

T °C or K temperature

V l culture volume in the bioreactor

X g l-1 biomass concentration

YX/S g g-1 yield of biomass per consumed glucose

µ h-1 specific growth rate

Indices

an in flux to anabolism

en in flux to energy metabolism

of flux to overflow metabolism

ox flux to oxidative m etabolism

NOMENCLATURE

44


45

ACKNOWLEDGEMENT

This work was partially financed by the Swedish centre for Bioprocess technology to which I would like to express my gratitude.

Professor Sven-Olof Enfors, to whom I am greatly thankful, supervised this work: Your deep knowledge in biotechnology and your enthusiasm has been a great inspiration.

Professor Lena Häggström for thoroughly reviewing this thesis in the last minute.

Ingrid Svensson, Ling Han and Gustav Silfversparre for their contributions in the presented articles. Jan Norling at AstraZeneca AB for the contribution on the solubility work and Cynthia Hassinen for the assisting during the mechanical stress cultivations. Cecilia Heyman and collegues at Albanova library, for rapid delivery of papers.

Associate professor Andres Veide, for his inspiring devotion in research in general and in beer-drinking in particular. Professor Gen Larsson, for handing over interesting articles at times. Tommy, for keeping me alive by giving me warnings of heat and electricity in the lab; Marita, for her effectiveness; Erik and Eric, for all computer support; Ela, for her warm personality and analytical competence; Bato, for the friendly jokes; Magda, for finally speaking Swedish; Katrin, Emma and Ling for lunches and fikas and beers. All the former colleagues at the department.

Erika, for being the devoted planner, always one step ahead of me concerning thesis aspects and party concepts. Helene, for being the organizer of fun stuff such as beer-brewing and “Fjellrypa”. But most of all for the enormous support they have given to me during the last few months. Without you there would not be a thesis to present!

Mamma and pappa, for all the baby-sitting, and for never letting me believe that there were things I simply could not do. Daniel, for being proud of me.

Janne, for all the support and for believing in me, more than myself sometimes. Also, for being an excellent “hemma-pappa” during the last period of time: I owe you and I love you!

Amanda and David, for being the rays of sunshine ✹ in my life.

ACKNOWLEDGEMENT

46


47

REFERENCES

Åkesson M, Hagander P, Axelsson J. 2001. Avoiding acetate accumulation in Escherichia coli cultures using feedback control of glucose feeding. Biotechnol Bioeng 73(3):221-230.

Andersen KB, von Meyenburg K. 1980. Are growth rates of Escherichia coli in batch cultures limited by respiration? J Bacteriol 144(1):114-123.

Andersson L, Strandberg L, Enfors S-O. 1996. Cell segregation and lysis have profound effects on the growth of Escherichia coli in high cell density fed-batch cultures. Biotechnol Prog 12:190-195.

Bauer S, White MD. 1976. Pilot scale exponential growth of Esherichia coli W to high cell concentration with temperature variation. Biotechnol bioeng 18:839-846.

Beveridge TJ. 1999. Structures of gram-negative cell-walls and their derived membrane vesicles. J Bacteriol 181(16):4725-4733.

Böck A, Sawers G. 1996. Fermentation. In: Neidhart FC, editor. Escherichia coli and Salmonella. Cellular and molecular biology. Washington, D.C.: ASM Press. p 262-282.

Boström M, Markland K, Sandén AM, Hedhammar M, Hober S, Larsson G. 2005. Effect of substrate feed on recombinant protein secretion, degradation and inclusion body formation in Escherichia coli. Appl Microbiol Biotechnol 68:82-90.

Brune G, Schoberth SM, Sahm H. 1982. Anaerobic treatment of an industrial wastewater containing acetic acid, furfural, and sulfite. Process Biochem 17(3):20-22, 24, 35.

Bucklin SE, Fujihara Y, Leeson MC, Morrison DC. 1994. Differential Antibiotic-induced release of endotoxin from gram-negative bacteria. Eur J Clin Microbiol 13(1):43-51.

Bull AT, Ward AC, Goodfellow M. 2000. Dearch and discovery strategies for biotechnology: the paradigm shift. Microbiol Mol Biol Rev 64(3):573-606.

Casadei MA, Manas P, Niven G, Needs E, Mackey BM. 2002. Role of membrane fluidity in pressure resistance of Escherichia coli NCTC 8164. Appl Env Microbiol 68(12):5965-5972.

Castan A, Näsman A, Enfors S-O. 2002. Oxygen enriched air supply in Escherichia coli processes: production of biomass and recombinant growth hormone. Enzyme Microb Tech 30:847-854.

Chaib F, Bernier A, Braendli E. 1995. Towards large scale methods for the selective release of periplasmic human cystatin C from Escherichia coli. Biotechnol Tech 9(3):179-184.

Chamsarta S, Hewitt CJ, Nienow AW. 2005. The impact of fluid mechanical stress on Cornynebacterium glutamicum during continuous cultivation in an agitated bioreactor. Biotechnol Lett 27:693-700.

Choi JH, Jeong KJ, Kim SC, Lee SY. 2000. Efficient secretory production of alkaline phosphatase by high cell density culture of recombinant Escherichia coli using Bacillus sp. endoxylanase signal sequence. Appl Microbiol Biotechnol 53:640-645.

Cronan JE, Gelmann EP. 1973. An Estimate of the minimum amount of unsaturated fatty acid required for growth of Escherichia coli. J Biol Chem 248(4):1188-1195.

de Maré L, Velut S, Ledung E, Cimander C, Norrman B, Nordberg Karlsson E, Holst O, Hagander P. 2005. A cultivation technique for E. coli fed-batch cultivations operating close to the maximum oxygen transfer capacity of the reactor. Biotechnol Lett 27:983-990.

Demain AL. 2000. Small bugs, big buisness: The economic power of the microbe. Biotechnol Adv 18:499-514. Doyle RJ, Chaloupka J, Vinter V. 1988. Turnover of cell walls in microorganisms. Microbiol rev 52(4):554-567. Edwards N, Beeton S, Bull AT, Merchuk JC. 1989. A novel device for the assessment of shear effects on suspended

microbial cultures. Appl Microbiol Biotechnol 30:190-195. El-Mansi M. 2004. Flux to acetate and lactate excretions in industrial fermentations: physiological and biochemical

implications. J Ind Microbiol Biotechnol. 31(7):295-300. Ellis RJ, Hartl FU. 1999. Principles of folding in the cellular environment. Curr opin struct biol 9:102-110. Enfors S-O, editor. 2004. Physiological Stress Responses in Bioprocesses. Berlin Heidelberg: Springer-Verlag. Etchegaray J-P, Jones PG, Inouye M. 1996. Differential thermoregulation of two highly homologous cold-shock genes, cspA

and cspB of Escherichia coli. Genes to cells 1:171-178. Fahnert B, Lilie H, Neubauer P. 2004. Inclusion bodies: Formation and utilization. Adv Biochem Engin/Biotechnol 89:93-

142. Fraenkel DG. 1996. Glycolysis. In: Neidhardt FC, editor. Escherichia coli and Salmonella. Cellular and molecular biology.

Washington, D.C.: ASM Press. p 189-198. Georgiou G, Shuler ML, Wilson DB. 1988. Release of periplasmic enzymes and other physiological effects of b-lactamase

overproduction in Escherichia coli. Biotechnol Bioeng 32(6):741-748. Georgiou G, Valax P. 1996. Expression of correctly folded proteins in Escherichia coli. Curr opin biotechnol 7:190-197. Han K, Lim HC, Hong J. 1992. Acetic acid formation in Escherichia coli fermentation. Biotechnol Bioeng 39(663-671). Han L, Enfors S-O, Häggström L. 2003. Escherichia coli high-cell-density culture: carbon mass balances and release of outer

membrane components. Bioproc Biosyst Eng 25:205-212. Hengge-Aronis R. 2002. Signal Transduction and Regulatory Mechanisms involved in control of the sigma s (RpoS) subunit

in Control of RNA polymerase. Microbiol Mol Biol Rev 66(3):373–395.

REFERENCES

48

Herbert D, Elsworth R, Telling RC. 1956. The continuous culture of bacteria; a theoretical and experimental study. J Gen Microbiol 14:601-622.

Herendeen SL, VanBogelen RA, Neidhardt FC. 1979. Levels of major proteins of Escherichia coli during growth at different termperatures. J Bacteriol 139(1):185-194.

Hewitt CJ, Boon LA, MacFarlane CM, Nienow AW. 1998. The use of flow cytometry to study the impact of fluid dynamical stress on Escherichia coli W3110 during continuous cultivation in an agitated bioreactor. Biotechnol Bioeng 59(5):612-620.

Hoekstra D, van der Laan JW, de Leij L, Witholt B. 1976. Release of outer membrane fragments from normally growing Escherichia coli. Biochim Biophys Acta 455:889-899.

Hoffmann F, Rinas U. 2004. Stress induced by recombinant protein production in Escherichia coli. In: Enfors S-O, editor. Physiological stress responses in bioprocesses. Berlin-Heidelberg: Springer.

Ingraham JL, Marr AG. 1996. Effect of temperature, pressure, pH, and osmotic stress on growth. In: Neidhart FC, editor. Escherichia coli and Salmonella. Cellular and molecular biology. Washington: ASM Press. p 1570-1578.

Irvin RT, MacAlister TJ, Chan R, Costerton JW. 1981. Citrate-tris(hydroxymethyl)aminomethane-mediated release of outer membrane sections from the cell envelope of a deep-rough (heptose defiecient lipopolysaccharide) strain of Escherichia coli O8. J Bacteriol 145(3):1386-1396.

Ishiguro EE, Vanderwel D, Kusser W. 1986. Control of Lipopolysaccharide Biosynthesis and release by Esherichia coli and Salmonella typhimurium. J Bacteriol 168(1):328-333.

Jahic M, Wållberg F, Bollok M, Garcia P, Enfors S-O. 2003. Temperature limited fed-batch technique for control of proteolysis in Pichia pastoris bioreactor cultures. Microbial Cell Fact 2(6).

Jiang W, Hou Y, Inouye M. 1997. CspA, the Major Cold-shock Protein of Escherichia coli, Is an RNA Chaperone. J Biolog Chem 272(1):196-202.

Johnston CJ, Finkelstein JN, Gelein R, Oberdorster G. 1998. Pulmonary cytokine and chemokine mRNA levels after inhalation of lipopolysaccharide in C57BL/6 mice. Toxicol Sci 46(2):300-307.

Jones PG, VanBogelen RA, Neidhardt FC. 1987. Induction of Proteins in Response to Low Temperature in Escherichia coli J Bacteriol 169(5):2092-2095.

Kadner RJ. 1996. Cytoplasmic membrane. In: Neidhart FC, editor. Escherichia coli and Salmonella. Cellular and molecular biology. Washington: ASM Press. p 58-87.

Kim Y-H, Han KY, Lee K, Lee J. 2005. Proteome response of Escherichia coli fer-batch culture to temperature downshift. Appl Microbiol Biotechnol 68:786-793.

Koch AL. 1998. The Biophysics of the gram-negative periplasmic space. Crit Rev Microbiol 24(1):23-59. Konstantinov KB, Nishio N, Yoshida T. 1990. Glucose feeding strategy accounting for the decreasing oxidative capacity of

recombinant Escherichia coli in fed-batch cultivation for phenylalanine production. J Ferment Bioengineer 70(4):253-260.

Laht T-M, Kask S, Elias P, Adamberg K, Paalme T. 2002. Role of arginine in the development of secondary microflora in Swiss-type cheese. Int Dairy J 12(10):831-840.

Larsson G, Pham H, Enfors S-O. 1990. The pH-auxostat as a tool for studying microbial dynamics in continuous fermentation. Biotechnol Bioeng 36:224-232.

Larsson TM, Lam H. 2004. Process Analytical Technology in biopharmaceutical production: Past successes and future challenges. J PAT 1(1):20-22.

Lee SY. 1996. High cell-density culture of Escherichia coli. TiBtech 14:98-105. Lin H, Hoffmann F, Rozkov A, Enfors S-O, Rinas U, Neubauer P. 2004. Change in extracellular cAMP concentration is a

sensitive reporter for bacterial fitness in high-cell-density cultures of Escherichia coli. Biotechnol Bioeng 87(5):602-613.

Loeb MR, Kilner J. 1978. Release of a special fraction of the outer membrane from both growing and phage T4-infected Escherichia coli B. Biochim Biophys Acta 518:117-127.

Luli GW, Strohl WR. 1990. Comparison of growth, acetate production, and acetate inhibition of Escherichia coli strains in batch and fed-batch fermentations. Appl Environ Microbiol 56(4):1004-10011.

MacBean RD, Hall RJ, Linklater PM. 1979. Analysis of pH-stat continuous cultivation and the stability of the mixed fermentation in continuous yogurt production. Biotechnol Bioeng 21(9):1517-1541.

Mackowiak PA. 1984. Relationship between growth temperature and shedding of lipopolysaccharides by gram-negative Bacilli. Eur J Clin Microbiol 3(5):406-410.

Magnusson LU, Farewell A, Nyström T. 2005. ppGpp: a global regulator in Escherichia coli. Trends Microbiol 13(5):236-242.

Martin GA, Hempfling WP. 1976. A method for the regulation of microbial population density during continous culture at high growth rates. Arch Microbiol 107:41-47.

Marvin HJ, ter Beest MB, Witholt B. 1989. Release of outer membrane fragments from wild-type Escherichia coli and from several E.coli lipopolysaccharide mutants by EDTA and heat shock treatments. J Bacteriol 171(10):5262-5267.

Mergulhao FJM, Summers DK, Monteiro GA. 2005. Recombinant protein secretion in Escherichia coli. Biotechnol Adv 23:177-202.

Meyer H-P, Leist C, Fiechter A. 1984. Acetate formation in continuous culture of Escherichia coli K12 D1 on defined and complex media. J Biotechnol 1:355-358.

Minkevich IG. 1986. The effect of automatic pH maintenance on steady state cultivation regimes and their stability. A theoretical study of the case of a perfect pH controller and perfect stirring. Acta Biotechnol 6(2):175-187.

Murby M, Uhlén M, Ståhl S. 1996. Upstream strategies to minimize proteolytic degradation upon rwecombinant production in Escherichia coli. Protein Express Purif 7:129-136.


49

Neidhardt FC, Ingraham JL, Schaechter M. 1990. Physiology of the bacterial cell, a molecular approach. Sunderland, Massachusetts: Sinauer Associates, Inc.

Nikaido H. 1996. Outer membrane. In: Neidhart FC, editor. Escherichia coli and Salmonella. Cellular and molecular biology. Washington: ASM Press. p 29-47.

Nikaido H. 2003. Molecular basis of bacterial outer membrane permeability revisited. Microbiol Mol Biol Rev 67(4):593-656.

Nyström T. 2004. Growth versus maintenance: a trade-off dictated by RAN polymerase availability and sigma factor competition? Mol Microbiol 54(4):855-862.

Oliver DB. 1996. Periplasm. In: Neidhart FC, editor. Escherichia coli and Salmonella cellular and molecular biology. Washington, D.C.: ASM press. p 88-103.

Park JT. 1996. The murein sacculus. In: Neidhart FC, editor. Escherichia coli and Salmonella. Cellular and molecular biology. Washington: ASM Press. p 48-57.

Pelltier C, Bourlioux P, van Heijenoort J. 1994. Effects of sub-minimal inhibitory concentrations of EDTA on growth of Escherichia coli and the release of lipopolysaccharide. FEMS Microbiol Lett 117(2):203-206.

Petsch D, Anspach FB. 2000. Endotoxin removal from protein solutions. J Biotechnol 76:97-119. Pham HTB, Larsson G, Enfors S-O. 1999. Modelling of aerobic growth of Saccharomyces cerevisiae in a pH-auxostat.

Bioprocess Eng 20(6):537-544. Phue J-N, Shiloach J. 2004. Transcription levels of key metabolic genes are the cause for different glucose utilization

pathways in E. coli B (BL21) and E. coli K (JM109). J Biotechnol 109:21-30. Postma P, Lengeler J, Jacobson G. 1996. Phosphoenolpyruvate: Carbohydrate Phosphotransferase Systems. In: Neidhardt

FC, editor. Escherichia coli and Salmonella. Cellular and molecular biology. 2 ed. Washington DC: ASM Press. p 1149-1174.

Raetz C. 1996. Bacterial Lipopolysaccharides: a remarkable family of bioactive macroamphipiles. In: Neidhart FC, editor. Escherichia coli and Salmonella. Cellular and molecular biology. Washington, D.C.: ASM Press. p 1035-1063.

Raivio TL. 2005. Envelope stress response and gram-negative bacterial pathogenesis. Mol Microbiol 56(5):1119-1128. Ramchuran SO. 2005. Bench-Scale production of heterologous proteins from extremophiles. Lund: Lund University, PhD-

thesis. Ramirez DM, Bentley WE. 1999. Characterization of stress and protein turnover from protein over-expression in fed-batch E.

coli cultures. J Biotechnol 71:39-58. Rhodius VA, Suh WC, Nonaka G, West J, Gross CA. 2006. Conserved and varible functions of the σE stress response related

genomes. PLOS Biology 4(1):43-59. Riesenberg D, Schulz V, Knorre WA, Pohl H-D, Korz D, Sanders EA, Ross A, Deckwer W-D. 1991. High cell density

cultivation of EScherichia coli at controlled specific growth rate. J Bacteriol 20:17-28. Rozkov A. 2001. Control of proteolysis of recombinant proteins in Escherichia coli. Stockholm: Royal Institute of

technology, PhD-thesis. Rozkov A, Enfors S-O. 2004. Analysis and control of proteolysis of recombinant proteins in Escherichia coli. In: Enfors S-O,

editor. Physiological stress responses in bioprocesses. Germany: Springer. Sahoo S, Verma RK, Suresh AK, Krishnamurthy Rao K, Bellare J, Suraishkumar GK. 2003. Macro-level and genetic-level

responses of Bacillus subtilis to shear stress. Biotechnol Prog 19:1689-1696. Sandén AM. 2004. Impact of glucose feed rate on productivity and recombinant quality in Escherichia coli. Stockholm:

Royal Institute of Technology, PhD-thesis. Satroutdinov AD, Christyakova TI, Dedyukhina EG, Kaparullina EN, Eroshin VK. 2005. Bacterial strain characterizing by

EDTA requirement. App Biochem Microbiol 41(5):469-473. Schein CH, Noteborn MHM. 1988. Formation of soluble recombinant proteins in Escherichia coli is favored by lower growth

temperature. Bio/Technology 6:291-294. Schletter J, Heine H, Ulmer AJ, Rietschel ET. 1995. Molecular mechanism of endotoxin activity. Arch Microbiol 165:383-

389. Shiloach J, Fass R. 2005. Growing E. coli to high cell density- A historical perspective on method development. Biotechnol

Adv 23:345-357. Shokri A, Sandén AM, Larsson G. 2003. Cell and process design for targeting of recombinant protein into the culture

medium of Escherichia coli. Appl Microbiol Biotechnol 60:654-664. Siano SA. 1999. Use of the abiotic proton balance for describing the pH-auxostat. Biotechnol Bioeng 64(6):755-758. Silfversparre G, Enfors S-O, Han L, Häggström L, Skogman H; 2002. Method for growth of microorganisms, minimising the

release of endotoxins from the bacteria to the surrounding medium. Simpson DR, Wiebe MG, Robson GD, Trinci APJ. 1995. Use of pH auxostats to grow filamentous fungi in continuous flow

culture at maximum specific growth rate. FEMS Microbiol Lett 126(2):151-158. Sinensky M. 1974. Homoeviscous adaptation. a homeostatic process that regulates the viscosity of membrane lipids in

Escherichia coli. Proc Nat Acad Sci USA 71(2):522-525. Sonnleitner B, Käppeli O. 1986. Growth of Saccharomyces cerevisae is controlled by its limited respiratory capacity:

Formulation and verification of a hypothesis. Biotechnol Bioeng 28:927-937. Sullivan KH, Hegeman GD, Cordes EH. 1979. Alteration of the fatty acid composition of Escherichia coli by growth in the

presence of normal alcohols. J Bacteriol 138(1):133-138. Tam C, Missiakas D. 2005. Changes in lipopolysaccharide structure induce the sE-dependent response of Escherichia coli.

Mol Microbiol 55(5):1403-1412. Teich A, Meyer S, Lin HY, Andersson L, Enfors S-O, Neubauer P. 1999. Growth rate dependent concentration changes of

the starvation response regulators sigma s and ppGpp in glucose-limited fed-batch and continuous cultures of Escherichia coli. Biotechnol Prog 15:123-129.

REFERENCES

50

Thomas CR. 1990. Problems of shear in biotechnology. In: Winkler MA, editor. Chemical engineering problems in biotechnology: Elsevier Applied Sciences, UK. p 23-94.

Toma MK, Ruklisha MP, Vanags JJ, Zeltina MO, Lelte MP, Galinine NI, Viesturs UE, Tengerdy RP. 1991. Inhibition of microbial growth and metabolism by excess turbulence. Biotechnol Bioeng 38:552-556.

Tsuchido T, Katsui N, Takeuchi A, Takano M, Shibasaki I. 1985. Destruction of the outer membrane permeability barrier of Escherichia coli by heat treatment. Appl Environ Microbiol 50(2):298-303.

Turner C, Gregory ME, Turner MK. 1994. A study of the effect of the specific growth rate and acetate on recombinant protein production of Escherichia coli JM107. Biotechnol Lett 16(9):891-896.

U.S Department of Health and Human Services, FDA, CDER, CVM, ORA. 2004. PAT -A framework for Innovative Pharmaceutical Development, Manufacturing, and Quality Assurance.

van den Berg B, Ellis RJ, Dobson CM. 1999. Effect of molecular crowding on protein folding and aggregation. EMBO J 18:6927-6933.

Vásquez-Laslop N, Lee H, Hu R, Neyfakh AA. 2001. Molecular sieve mechanism of selective release of cytoplasmic proteins by osmotically shocked Escherichia coli. J Bacteriol 183(8):2399-2404.

Verduyn C, Postma E, Scheffers WA, van Dijken JP. 1992. Effect of benzoic acid on metabolic fluxes in yeast: A continuous culture study on the regulation of respiration and alcoholic fermentation. Yeast 8:501-517.

Verschuren PG, Van der Heuvel JC. 2002. Susbstrate controlled development of anaerobic acidifying aggregates at different shear rates in a gas lift reactor. Biotechnol Bioeng 77(3):306-315.

Vorachek-Warren MK, Carty SM, Lin S, Cotter RJ, Raetz CRH. 2002. An Escherichia coli mutant lacking the cold shock-induced palmitoleoyltransferase of lipid A biosynthesis. J Biolog Chem 277(16):14186-14193.

Wensink J, Witholt B. 1981. Outer-membrane vesicles released by normally growing Escherichia coli contain very little lipoprotein. Eur J Biochem 116:331-335.

Xu B, Jahic M, Enfors S-O. 1999. Modeling of overflow metabolism in batch and fed-batch cultures of Escherichia coli. Biotechnol Prog 15(1):81-90.

Yamané T, Shimizu S. 1984. Fed-batch techniques in microbial propcesses. Adv Biochem Eng 30:148-192. Yu ST, Lin CC, Too JR. 2005. PHBV production by Ralstonia eutropha in a continuous stirred tank reactor. Process

Biochem 40(8):2729-2734. Zhao Y, Lin Y-H. 2002. Distribution of ATP and reducing equivalents in Corynebacterium glutamicum during amino acid

resolution. Process Biochem 37(12):1455-1461. Zhou L, Srisatjaluk R, Justus DE, Doyle RJ. 1998. On the origin of membrane vesicles in Gram-negative bacteria. FEMS

Microbiol Lett 163:223-228.

Documents

THE TEMPERATURE LIMITED FED BATCH TECHNIQUE FOR10943/FULLTEXT01.pdf · THE TLFB TECHNIQUE FOR CONTROL OF E. COLI CULTURES 1 INTRODUCTION Biotechnology has been a major participant