Impact of Glucose Uptake Rate on Recombinant Protein Production in Escherichia Coli

Impact of glucose uptake rate on recombinant protein production in Escherichia coli

Emma Bäcklund M. Sc.

Royal Institute of Technology

Stockholm 2011

School of Biotechnology Royal Institute of Technology SE-106 91 Stockholm Sweden

Printed at AJ E-print AB Oxtorgsgatan 9 SE-111 57 Stockholm Sweden

ISBN 978-91-7415-994-3 ISSN 1654-2312 TRITA-BIO Report 2011:18

©Emma Bäcklund (2011). Impact of glucose uptake rate on recombinant protein production in Escherichia coli. School of Biotechnology, Royal Institute of Technology (KTH), Albanova University Center, Stockholm, Sweden.

ABSTRACT Escherichia coli (E. coli) is an attractive host for production of recombinant proteins, since it generally provides a rapid and economical means to achieve high product quantities. In this thesis, the impact of the glucose uptake rate on the production of recombinant proteins was studied, aiming at improving and optimising production of recombinant proteins in E. coli.

E. coli can be cultivated to high cell densities in bioreactors by applying the fed-batch technique, which offers a means to control the glucose uptake rate. One objective of this study was to find a method for control of the glucose uptake rate in small-scale cultivation, such as microtitre plates and shake flasks. Strains with mutations in the phosphotransferase system (PTS) where used for this purpose. The mutants had lower uptake rates of glucose, resulting in lower growth rates and lower accumulation of acetic acid in comparison to the wild type. By using the mutants in batch cultivations, the formation of acetic acid to levels detrimental to cell growth could be avoided, and ten times higher cell density was reached. Thus, the use of the mutant strains represent a novel, simple alternative to fed-batch cultures.

The PTS mutants were applied for production of integral membrane proteins in order to investigate if the reduced glucose uptake rate of the mutants was beneficial for their production. The mutants were able to produce three out of five integral membrane proteins that were not possible to produce by the wild-type strain. The expression level of one selected membrane protein was increased when using the mutants and the expression level appeared to be a function of strain, glucose uptake rate and acetic acid accumulation.

For production purposes, it is not uncommon that the recombinant proteins are secreted to the E. coli periplasm. However, one drawback with secretion is the undesired leakage of periplasmic products to the medium. The leakage of the product to the medium was studied as a function of the feed rate of glucose in fed-batch cultivations and they were found to correlate. It was also shown that the amount of outer membrane proteins was affected by the feed rate of glucose and by secretion of a recombinant product to the periplasm.

The cell surface is another compartment where recombinant proteins can be expressed. Surface display of proteins is a potentially attractive production strategy since it offers a simple purification scheme and possibilities for on-cell protein characterisation, and may in some cases also be the only viable option. The AIDA-autotransporter was applied for surface display of the Z domain of staphylococcal protein A under control of the aidA promoter. Z was expressed in an active form and was accessible to the medium. Expression was favoured by growth in minimal medium and it seemed likely that expression was higher at higher feed rates of glucose during fed-batch cultivation. A repetitive batch process was developed, where relatively high cell densities were achieved whilst maintaining a high expression level of Z.

Keywords: AIDA-autotransporter, Escherichia coli, fed-batch, glucose uptake rate, integral membrane proteins, outer membrane proteins, periplasmic retention, phosphotransferase system, recombinant proteins, specific growth rate, surface expression.

LIST OF PUBLICATIONS

The thesis is based on the following papers, referred to in the text by their Roman numerals:

I. Bäcklund E, Markland K, Larsson G (2008). Cell engineering of Escherichia coli allows high cell density accumulation without fed-batch process control. Bioprocess and Biosystems Engineering 31:11-20

II. Bäcklund E, Reeks D, Markland K, Weir N, Bowering L, Larsson G (2008). Fedbatch design for periplasmic product retention in Escherichia coli. Journal of Biotechnology 135:358-365

III. Bäcklund E, Ignatuschenko M, Larsson G (2011). Suppressing glucose uptake and acetic acid production increases membrane protein overexpression in Escherichia coli. Accepted for publication in Microbial Cell Factories.

IV. Gustavsson M, Bäcklund E, Larsson G (2011). Optimisation of surface expression using the AIDA autotransporter. Manuscript.

CONTENTS

1 INTRODUCTION ...... 1 1.1 ADAPTIVE RESPONSES TO CHANGES IN GROWTH RATE ...... 3 1.2 THE MEMBRANE STRUCTURE OF E. COLI ...... 4 1.3 CONTROL OF GLUCOSE UPTAKE RATE ...... 6 1.3.1 Cultivation techniques ...... 6 1.3.2 Glucose uptake ...... 8 1.3.3 Acetic acid formation – a result of high glucose uptake rate ...... 10 1.3.4 Reduction of acetate formation ...... 11 1.4 LIMITING FACTORS IN RECOMBINANT PROTEIN PRODUCTION ...... 12 1.4.1 Cytoplasmic production ...... 14 1.4.2 Periplasmic production ...... 21 1.4.3 Production in the inner membrane ...... 25 1.4.4 Surface display of proteins in E. coli ...... 27 1.4.5 Extracellular production ...... 33 2 PRESENT INVESTIGATION ...... 34 2.1 A CELLULAR ALTERNATIVE TO FED-BATCH CULTURES (I, III) ...... 35 2.1.1 Strain evaluation (I) ...... 35 2.1.2 Production of integral membrane proteins by the PTS-mutants (III) ...... 41 2.2 IMPACT OF FEED-RATE ON PROCESSES WITH OTHER PRODUCT LOCALISATIONS THAN THE CYTOPLASM (II, IV) ...... 46 2.2.1 Leakage of periplasmic products in relation to the glucose uptake rate (II) ...... 46 2.2.2 Optimisation of surface expression using the AIDA autotransporter (IV) ...... 52 3 CONCLUDING REMARKS ...... 58 4 ABBREVIATIONS ...... 62 5 ACKNOWLEDGMENT ...... 64 6 REFERENCES ...... 65

1 INTRODUCTION

Since the first announcement on microbial production of a protein of human origin, insulin, was made in 1978 by researchers at the company Genentech (Genentech, press release, 1978), recombinant protein production has become a routine business. Today, it is a common strategy for production of many high-value proteins used in for example various medical applications (e.g., vaccines, recombinant factor VIII for treatment of haemophilia, insulin for treatment of diabetes and tissue-plasminogen activator against stroke). The principle behind this technology is relatively simple: a foreign gene, encoding a target protein of interest, can be introduced into a host cell that will use its own cellular machinery to translate the gene into the desired protein product. However, this important breakthrough would never have been realised without the pioneering work made in the early seventies on how to isolate and amplify genes (or DNA) and then insert them into specific genetic locations to create transgenic organisms (Cohen, et al., 1973; Lobban and Kaiser, 1973; Morrow, et al., 1974), hence forming the ground for what we today know as recombinant DNA technology. Although Escherichia coli (E. coli) was used for expression of insulin in this first example, the use of host cells is not restricted to bacteria such as E. coli but many other prokaryotic (e.g., Bacillus) and eukaryotic (yeast, insect and mammalian) cells can be used as well. Regardless of the specific host cell type used, the productivity is influenced by environmental conditions (e.g., temperature, pH, availability of oxygen and nutrients) as well as by genetic factors (e.g., the gene/protein itself, the promoter strength, mRNA stability, codon usage, gene copy numbers, availability of co-factors and various helper systems that aid in the expression). Thus, in the end, process optimisation boils down to finding the appropriate combination of environmental conditions and genetic factors that will result in the highest amount of active protein.

The focus when producing recombinant proteins in the pharmaceutical industry today is generally either on i) large-scale production of particular commercial protein products or on ii) small scale high-throughput production (HTP) of proteins that are used either for structural and functional studies or for high-throughput screening (HTS) against potential drug leads in the drug development process. E. coli, with its ability to grow rapidly to high cell densities on inexpensive substrates, its well-

1 characterised genetics and the availability of numerous cloning vectors and mutant host strains, is an attractive host for production of recombinant proteins (Schmidt, 2004).

The goal for the large-scale industrial production of a protein is to achieve a high total productivity in order to produce a large amount of the protein in a cost efficient manner. Operator supervised bioreactors, which offer a high level of control and regulation of the process are used and parameters like temperature, pH and dissolved oxygen tension (DOT) are measured and regulated. The fed-batch technique is usually applied, where a glucose feed is continuously added to the bioreactor during the cultivation, making it possible to control the glucose uptake rate of the cells. The fed- batch technique enables the establishment of high cell densities, since growth rate and by-product formation are controlled.

The goal for high-throughput production of proteins is to achieve “sufficient” amounts of soluble proteins for functional or structural studies. This production usually relies on unsupervised batch cultivation in small scale, most commonly micro titre plates or shake flasks. The cells grow exponentially until limitations related to e.g., oxygen availability or by-product formation arises. The level of control of the environmental conditions is low in this type of small scale shaken systems and the conditions changes rapidly as a consequence of the exponential biomass increase. For practical reasons, the fed-batch technology is not applicable for control of the glucose uptake rate in such systems, and there consequently is a need for other innovative strategies for controlling the glucose uptake rate; especially as the glucose uptake rate has been shown to have an impact on product-associated parameters such as specific productivity, solubility and proteolysis of the recombinant protein in earlier studies (Boström, et al., 2005; Ryan, et al., 1996; Sandén, et al., 2005; Sandén, et al., 2002).

In general, recombinant proteins are preferably produced in active, soluble forms. One problem with production of recombinant proteins in E. coli is that the overexpressed proteins are not always properly folded and then associate into mainly non-active and insoluble aggregates, which are termed inclusion bodies. Proteins in inclusion bodies may regain activity after a refolding process. Refolding is, however, a complicated

2 process, and the process has to be optimised for each protein in question (Hauke, et al., 1998) and is therefore not an applicable strategy in HTP applications. However, some proteins fold properly if they are secreted to the periplasm. Other “difficult-to- express” proteins such as membrane proteins and toxic proteins are also preferably transported to other parts of the cell than the cytoplasm. In order to understand the impact of the glucose uptake rate on these kinds of processes where the product is localised to other parts of the cell than the cytoplasm, further investigations are needed.

1.1 Adaptive responses to changes in growth rate

The growth rate of cells varies depending on the growth conditions, i.e. the growth rate is influenced by factors such as substrate concentration, growth medium, temperature, pH and the supply of oxygen. In general, fast growing cells contain more DNA, RNA, ribosomes, proteins, phospholipids and cell wall material, and tend to increase in size (Lengeler and Postma, 1999). Cells respond to carbon or amino acid limitation by inhibited RNA and protein synthesis. Also the DNA replication as well as the biosynthesis of carbohydrates, phospholipids and cell wall constituents is inhibited and the cell size decreases. This set of responses, with a tight coupling between growth rate, ribosomal synthesis and cell size is referred to as the stringent response and is mediated by the production of the alarmone guanosine tetraphosphate (ppGpp). E. coli uses two different pathways to produce ppGpp. The lack of amino acids results in the binding of uncharged tRNA to the ribosome, which activates the RelA enzyme leading to the formation of ppGpp. The lack of carbon, on the other hand, leads to the activation of the alternative pathway for ppGpp formation involving SpoT (Lengeler and Postma, 1999). The ppGpp bind to the RNA polymerase core enzyme, which affects the expression of a plethora of genes. In general, genes involved in cell proliferation and growth are negatively regulated by ppGpp, whereas genes involved in maintenance and stress defence are positively regulated (Magnusson, et al., 2005). The starvation sigma subunit, σS, accumulates in the cell whenever the growth rate is lowered and not only when the growth ceases. The synthesis of σS is positively controlled by ppGpp (Booth, 1999).

3 1.2 The membrane structure of E. coli

The lipid structure of the membranes depends on the glucose uptake rate (Shokri, et al., 2002). However, before discussing this dependency, an overview of the basic structures of the membranes will be given. The basic unit of a membrane is a bilayer that is formed by phospholipids organized in two layers with their polar head groups along the two surfaces and the acyl chains forming the nonpolar domain between. Membranes are dynamic with movement both across and in the plane of the bilayer. The bilayer serves as matrix and support for many proteins that are involved in transmembrane processes, including translocation of proteins and other molecules across membranes (Dowhan, 1997). “The fluid mosaic model” (Singer and Nicolson, 1972) has been used for many years for describing the nature of the membranes. The membranes proteins are in this model more or less viewed as icebergs floating in a sea of lipids. However, in the last years, this view has been shifting and the importance of transient, specialized regions called membrane rafts, which are enriched in special lipids or proteins has become clear (Luckey, 2008).

Pore

LPS Phospholipid

Outer Membrane

Lipoprotein

Periplasm

Peptidoglycan

Protein

Inner Membrane

Figure 1. Model of Escherichia coli cell envelope. The cell has a two-membrane structure composed of the cytoplasmic and the outer membrane. The periplasm, the space between theses membranes, contains the cell wall made of peptidoglycans.

4 The cell envelope of gram-negative bacteria, to which E. coli belongs, is a two- membrane structure of cytoplasmic and outer membrane (Fig 1). The space between the membranes is the periplasm where a thin cell wall consisting of peptidoglycan is situated. The cell wall gives shape and rigidity to the cell, and prevents the cell from lysing in dilute environments. The outer membrane contains lipopolysaccharide (LPS), a structure that is not found in the cytoplasmic membrane.

Escherichia coli membranes contain three major phospholipids: phosphatidylethanolamine (PE), phosphatidylglycerol (PG) and cardiolipin (CL). PE, which is the major phospholipid, constitutes roughly 75 % of the total phospholipid content, PG 18% and CL 5 % (Cronan and O.Rock, 1996). The phospholipid composition of the cytoplasmic and outer membrane is similar, but with a slight enrichment of PE in the outer membrane (Nikaido, 1996). PE is zwitterionic and does not carry a net charge at physiological pH, while PG and CL are anionic. Phospholipids contain both saturated and unsaturated fatty acids as well as cyclic fatty acids, which are formed by methylation of unsaturated ones. E. coli adjusts the fatty acid composition of its phospholipids in response to growth temperature in order to preserve a more or less constant degree of membrane fluidity. The proportion of unsaturated fatty acids increases as the temperature decreases and vice versa for the saturated ones. This change results in more or less constant fluidity of the membranes since the melting point of lipids decreases as the proportion of unsaturated fatty acids increases (Neidhardt, et al., 1990). The total amount of phospholipids does not vary with the growth rate but the composition of the phospholipids does (Cronan and O.Rock, 1996). PG reaches a maximum at a specific growth rate of 0.3 h-1 (Shokri, et al., 2002) and this growth rate is associated with a high permeability of the membrane. The membrane becomes more rigid as the growth rate declines (Shokri, et al., 2003).

5 1.3 Control of glucose uptake rate

1.3.1 Cultivation techniques

Batch In a batch process all substrate components are available at high enough concentrations to make the reaction rate unrestricted with respect to substrate concentration (Fig 2). The biomass concentration increases exponentially, i.e. the specific growth rate (μ) is constant at μmax, until some factor e.g., by-product concentration, oxygen supply or low substrate concentration reduces the specific growth rate.

Oxygen limitation and the formation of the by-product acetic acid limit the usefulness of the batch technique. The batch technique is however the easiest cultivation method and it is used when proteins are produced either for functional and structural studies or for HTS. Further, screening for new production methods in the industrial process development is usually performed in batch mode.

!"#$%#

&'()*# !+,-.'()*#

Figure 2. The difference between batch and fed-batch cultivation. -1 F= feed rate of limiting substrate (l h ), Si=concentration of the substrate (g l-1)

Fed-batch The fed-batch technology is a common industrial method for recombinant protein production. One substrate component, usually glucose, is added to the bioreactor at a rate so that its concentration is growth rate limiting and thus the growth rate can be controlled via the feed (Fig 2). The condition for growth rate limitation in a fed-batch process is:

F/V(t)*Si < qs,max X(t)

-1 -1 Where F (l h ) is the feed rate, V (l) the cultivation volume, Si (g l ) the concentration -1 -1 of the substrate in the feed solution, qs,max (g g h ) the maximal specific consumption rate of the limiting substrate and X (g l-1) the cell mass at that time point (Enfors and Häggström, 2000). There are two main reasons to apply the fed-batch technique. First, the substrate limitation offers a tool for reaction rate control to avoid engineering limitations with respect to cooling and oxygen transfer. Secondly, the substrate limitation also permits a sort of metabolic control by which overflow metabolism, resulting in formation of acetic acid, can be avoided (Enfors and Häggström, 2000). The fed-batch technique makes is possible to obtain high cell densities and thus to reach a high total productivity.

The feed profile can be designed in different ways. A constant feed results in a continuously decreasing growth rate of the cells since the amount of substrate per cell decreases as a function of time. An exponential feed results in a constant growth rate of the cells, since each cell receives the same amount of substrate during the whole cultivation. By using exponential feed-profiles, exponential growth can be maintained, but at a lower rate than the maximal rate (μmax). A steady state with respect to substrate concentration and growth rate is established in exponential fed- batch cultures.

Fed-batch cultivations may be started as batch cultures and the feed of substrate is then started when the initial substrate is depleted. In the industry, however, it is common to start the feed of substrate directly after the inoculation of the reactor. Usually, an exponential feed is used until the oxygen limitation of the reactor is reached. To further increase the biomass, a constant feed is then applied, which results in gradually reduced specific growth rates of the cells.

7 1.3.2 Glucose uptake

Diffusion of carbohydrates through the outer membrane occurs mainly through the outer membrane channel forming proteins OmpC, OmpF and LamB (Nikaido, 1996). The phosphotransferase system (PTS) transports carbohydrates such as glucose, fructose and mannose from the periplasm to the cytoplasm. The PTS system is composed of both soluble proteins and proteins that are integrated in the cytoplasmic membrane (Fig 3). Enzyme I (EI) and phosphohistidine carrier protein (HPr) are general PTS proteins, while the enzymes IIs (EIIs) are sugar specific. EIIGlc is specific for glucose and EIIMan for mannose. At least one of the EII-domains is bound to the membrane. The glucose EII consists of the parts IIAGlc and IICBGlc where the latter is membrane bound, while the mannose EII consists of the IIABMan and the membrane bound parts IICMan and IIDMan (Fig 3). A series of reactions are involved when the sugar molecule enters the cytoplasm. A phosphate group is transferred from phosphoenolpyruvate (PEP) to EI and further to HPr. The phosphate group is then transferred to the EIIA domains and further to the EIIB/EIICB domains. These domains perform phosphorylation of the incoming sugar molecule (Postma, et al., 1996).

Figure 3. A model for PTS mediated uptake of carbohydrates. The scheme shows the general enzymes of the PTS: Enzyme I (EI), Phosphohistidine carrier protein (HPr) and two Enzymes II (EII). P indicates phosphorylation of the various enzymes and the sugar molecules. Modified from (Postma et al. 1996).

8 Glucose is transported over the cytoplasmic membrane mainly through the glucose and mannose specific PTS. GalP and the Mgl-system, proteins that are normally involved in the transport of galactose, may however also transport glucose into the cytoplasm. These proteins can be induced and used for glucose transport during glucose limiting conditions. Glucose that is transported into the cytoplasm by the Mgl-system or GalP is phosphorylated by glucokinase, encoded by the gene glk (Gosset, 2005).

Glucose uptake is also controlled by the repressor protein Mlc. The phoshorylated form of enzyme IICBGlc dominates in the absence of glucose. In this situation Mlc binds upstream of the ptsG promoter and works as a repressor. Addition of glucose to the medium results in dephosphorylation of enzyme IICBGlc. Mlc binds to the dephoshorylated enzyme IICBGlc and is thereby sequestered away from the operator making transcription of ptsG possible. Mlc is, as described above, thus capable of binding both DNA and proteins (Plumbridge, 2002).

The PTS system also has an important role in catabolite repression, which results in inhibition of gene expression and/or protein activity by the presence of a rapidly metabolisable carbon source in the medium. Glucose is the preferred carbon source for E. coli and decreased concentration of glucose in the medium results in accumulation of the phoshorylated form of enzyme IIAGlc and IICBGlc. The phosphorylated form of enzyme IIAGlc activates adenylate cyclase (AC), the enzyme that converts ATP to cAMP, resulting in increased levels of cAMP. The cAMP binds the product of the crp locus, termed the cAMP receptor protein (CRP). The cAMP- CRP complex causes the induction of catabolite-repressed genes allowing uptake of other sugars. The uptake of substrates is also regulated by another mechanism. The unphosphorylated form of IIAGlc dominates when the glucose concentration in the medium is high. When enzyme IIAGlc is in the unphosphorylated form, it can inactivate several transport systems for non-PTS carbon sources, by binding to the transporters. This mechanism is called inducer exclusion (Lengeler and Postma, 1999).

9 1.3.3 Acetic acid formation – a result of high glucose uptake rate

Aerobic growth of E. coli on excess of glucose leads to the excretion of partially oxidised metabolites, mostly in the form of acetic acid. This phenomenon is called overflow metabolism and occurs if the glucose concentration exceeds a critical value. This value is approximately 20-30 mg l-1 glucose and corresponds to a specific growth rate of approximately 0.30 h-1 in minimal medium (Enfors and Häggström, 2000; Meyer, et al., 1984). The reason for the overflow metabolism is not clear but it may be a result of an imbalance between the glycolysis and the TCA-cycle or of saturation of the TCA-cycle or the electron transport chain (Lee, 1996).

The main route for acetate production is from Acetyl-CoA through the enzymes phosphotransacetylase (Pta) and acetatekinase (Ack) (Fig 4). This route generates ATP. The other route for acetate production is directly from pyruvate by pyruvate oxidase B (PoxB), but it plays a minor role (Wolfe, 2005).

Glucose!

PEP!

Pyruvate!

CO ! poxB 2 !

CO2! NADH! Acetyl-CoA! pta! Acetyl-P! ack! Acetate! ATP!

Figure 4. Simplified view of acetate formation in E. coli. Acetate is formed either by the Pta-Ack route or by the PoxB route.

Most E. coli strains have the ability to assimilate acetate. This is primarily done by the enzyme AMP-forming acetyl CoA-synthetase (AMP-ACS) (Wolfe, 2005). This route is utilized when acetate is the carbon source or when there is a need to reabsorb acetate that has been excreted as a consequence of overflow metabolism (Wolfe, 2005).

10 Acetate production is undesirable in recombinant protein production since it reduces bacterial growth even at concentrations as low as 0.5 g l-1 (Nakano, et al., 1997). Further, acetate formation reduces the yield of biomass and the production of recombinant proteins has also been shown to be negatively affected (Jensen and Carlsen, 1990; Turner, et al., 1994).

Acetate production depends on the bacterial strain, and it is known that E. coli B- strains do not accumulate as much acetic acid as E. coli K12 strains. It has been suggested that B-strains have a more active glyoxylate shunt than K12-strains, which results in the direction of the pyruvate into biosynthetic precursors, e.g., succinate and in the reduction of acetic acid (Phue and Shiloach, 2004; van de Walle and Shiloach, 1998). The lower accumulation of acetic acid in B-strains has also been explained by differences in the transcription level of the acetate production genes (Phue and Shiloach, 2004).

1.3.4 Reduction of acetate formation

The most commonly used approach in order to decrease the acetate production is the fed-batch technique. Another strategy is to use genetic modification. The targets genes are then i) genes that code for proteins involved in the uptake of glucose or ii) genes that code for enzymes that are active in pathways resulting in acetic acid.

Using an E. coli strain with a mutation in ptsG, the gene coding for EIIGlc, resulted in 20-40% decrease in the specific acetate yield in comparison to the wild type. Other consequences of the mutation were an increased biomass, from 13 to 19 g l-1 and an increase in recombinant protein concentration exceeding 50% (Chou, et al., 1994). Mlc overproducing strains were constructed by mutations in the promoter region of the chromosomal mlc gene. This strain showed a reduced acetate accumulation, and as a consequence of this, the pH of the media was higher in comparison to the wild type.

The effect of the mutation was also observed as an increase in the OD600 from 5 to 8 in comparison to the wild type (Cho, et al., 2005).

11 Another approach that has been used in order to decrease the accumulation of acetic acid is to replace the PTS system by the galactose permease. The galP gene is normally repressed when E. coli grows on glucose. Its transcription is however increased in PTS mutant strains. In this study, the chromosomal galP promoter was replaced by the trc promoter, making induction with IPTG (isopropyl β-D- thiogalactopyranoside) possible. The acetate concentration was decreased from 2,8 to 0,39 g l -1 and the GFP formation was increased four times in comparison to the wild type (De Anda, et al., 2006).

Mutant strains, which lack the enzymes that form acetic acid, have also been constructed and the excretion of acetate in these strains has been reduced (Bauer, et al., 1990). Recently a triple BL21 mutant lacking all known acetate production genes (ackA-pta-poxB) was constructed. Interestingly, this strain still accumulates acetic acid, which indicates that there are other alternative acetate production pathways in the cell (Phue, et al., 2010).

1.4 Limiting factors in recombinant protein production

The maintenance of a plasmid and high-level expression of the target protein represents a metabolic burden to the cell. Production of a recombinant protein can trigger stress responses in the cell that often resembles the cellular responses to environmental stress such as heat shock and stringent response (Sørensen and Mortensen, 2005). The extent of the stress responses is determined by the rates of transcription and translation, but stress can also emerge from the specific properties of the recombinant protein, i.e. misfolding, which results in the degradation of the recombinant product and in inclusion body formation (Hoffmann and Rinas, 2004). Production of recombinant proteins can also lead to ribosome destruction (Dong, et al., 1995) and in severe cases to cell-death (Miroux and Walker, 1996).

Generally, the goal in recombinant protein production is not only to maximise the amount of recombinant protein but also to achieve a protein with a high quality, i.e. a protein in an active, soluble and pure form. However, some proteins cannot be produced as active, soluble entities in the cytoplasm and must therefore be directed to other compartments of the cell. The advantages and disadvantages associated with

12 expression in each compartment are summarised in table 1. Overexpression in the outer membrane has the purpose of producing whole cells with recombinant proteins expressed at the surface, and differs in that context from the production in the other compartments, where the goal is to produce recombinant proteins for isolation and purification.

Table 1. Advantages and disadvantages with production of recombinant proteins in different compartments of the cell.

Compartment Advantages Disadvantages

Cytosol (soluble protein) Higher protein yield No S-S bond formation More complex purification Many proteases

Cytosol (inclusion High protein yield Solubilisation and refolding needed bodies) Protection from proteases Lower yield Reduced toxicity Higher cost Inclusion bodies are easy to isolate Refolding not always possible

Inner membrane (limited The product can be purified after Low yield to production of detergent extraction Limited space in the membrane membrane proteins) Accumulation of product may be toxic to the cell

Periplasm Fewer proteases Inclusion bodies may form Formation of S-S bonds Inefficient transport N-terminus authenticity Simpler purification if selective release of product possible

Outer membrane (limited On-cell protein characterisation possible Limited space in the membrane to surface expression of No purification of protein is needed Transport not always possible proteins) Increased stability of product Whole cells can simply be removed

Extracellular Less extensive proteolysis Transport not always possible Easier purification Diluted product N-terminus authenticity Reduced toxicity

13 1.4.1 Cytoplasmic production

The goal for recombinant protein production in the cytoplasm is either to obtain soluble proteins or to direct the product into inclusion bodies. The advantage with the cytoplasmic production is the high yield of product. The cytoplasm is a reducing environment, inhibiting the formation of disulphide bridges in the protein, a problem that might be circumvented by secretion to the periplasm. Other disadvantages associated with cytoplasmic production is the need for cell disruption, the complex purification from a mixture of endogenous proteins and the higher amounts of proteases in comparison to the periplasm. In some cases, cytoplasmic overexpression of a gene results in the presence of one extra N-terminal methionine in the product (Adams, 1968; Moerschell, et al., 1990), which may have negative impact on the stability and solubility of the product (Chaudhiri, et al., 1999). That the product ends up as inclusion bodies may for proteins that are easy to refold actually be beneficial. Proteins in inclusion bodies are easy to separate form the cell debris after cell disruption, they are mostly inactive and therefore not toxic to the cell, and they are generally protected from degradation by proteases and are relatively pure (Jürgen, et al., 2010; Makrides, 1996). However, when using this strategy, in vitro refolding is needed to achieve a correctly folded product. There is however no guarantee that proteins in inclusion bodies will regain activity or that the refolding will result in a high yield. Refolding and purification from inclusion bodies is usually more expensive and time consuming than purification of soluble proteins (Sorensen and Mortensen, 2005). The renaturation conditions, such as temperature, buffer, pH and ionic strength must be optimized for every protein (Hauke, et al., 1998) and using the refolding strategy in HTP applications is therefore not an option.

Folding The newly synthesized protein either folds independently or needs to be assisted by molecular chaperones to attain a correct and biologically active structure. The molecular chaperones are ”proteins that help the folding of other proteins, usually through cycles of binding and release, without forming part of their final structure” (Young, et al., 2004). The chaperones favor on-pathway folding by shielding interactive surfaces from each other as well as from the solvent, and by accelerating rate-limiting steps. Chaperones are also invaluable in protein secretion and

14 translocation where their function is to prevent folding of the target protein, thereby keeping them in a translocation-competent state. Chaperones are also involved in the process of disaggregation of already formed aggregates (Baneyx and Mujacic, 2004).

Figure 5. Chaperone-assisted folding in the cytoplasm. The nascent polypeptide first encounters TF or DnaK/DnaJ as it exits the ribosome. After release, the folding intermediate either reaches its native conformation or is transferred to GroEL/GroES for further folding assistance. Partially folded proteins may also be stabilized by the holding chaperones IbpA/B, Hsp31 and Hsp33 until folding chaperones are available. The disaggregating chaperone ClpB promotes disaggregation of unfolded proteins and cooperates with the folding chaperones to reactive them. 1

The molecular chaperones can be divided into three different groups based on their function, the folding (e.g., DnaK and GroEL), the holding (e.g., IbpB) and the disaggregating (e.g., ClpB) chaperones (Fig 5). Foldases constitute another important group of chaperones and their role is to accelerate rate-limiting steps in the folding

1 Reprinted by permission from Macmillan Publishers Ltd: [Nature Biotechnology], (Francois Baneyx and Mirna Mujacic, 2004, Recombinant protein folding and misfolding in Escherichia coli, Nature Biotechnology, Vol. 22, No 11, p 1399-1408), Copyright (2004)

15 pathways. The peptidyl-prolyl isomerases (PPIases) increases the rate of cis-trans isomerization of peptide bonds involving proline residues and the thiol disulfide oxidoreductases, known as the Dsb proteins, catalyse the formation and shuffling of disulphide bridges in the periplasm (Baneyx and Mujacic, 2004).

The heat shock response in E. coli is regulated by σ32 and it is induced by a large variety of stress factors including temperature shift, metabolic harmful substances and protein misfolding and aggregation. Major heat shock proteins are proteases and molecular chaperones including DnaK/DnaJ/GrpE and GroEL/GroES (Arséne, et al., 2000).

Proteolysis Proteolysis and folding are tightly connected processes in the cell. The degradation of misfolded proteins guarantees that abnormal polypeptides do not accumulate within the cell and conserves the cellular resources by the recycling of amino acids. The ATP-dependent cytoplasmic protease Lon and the Clp family of proteases are induced by heat shock (Gross, 1996).

One strategy that can be used in order to avoid degradation of the recombinant product is utilizing protease deficient mutants. Using such mutants is however often associated with reductions of the growth rate and it is not known if the cell compensates for the loss of one protease by increasing the concentration of another (Martínez-Alonso, et al., 2010). One exception is however the strain BL21 that is deficient of both Lon and OmpT but still exhibits good growth characteristics (Sørensen and Mortensen, 2005). However, in strains devoid of the Lon protease, the aggregation of over-produced proteins has shown to be enhanced (Corchero, et al., 1996). For proteins that are easy to refold, the deposition of them into inclusion bodies can be an alternative superior strategy for avoiding degradation from proteases (Hauke, et al., 1998).

16 Recovery and degradation of aggregated proteins Protein aggregates accumulate when the cell’s capacity for folding, unfolding and degradation is exceeded, which is usually a consequence of stress or of overexpression of recombinant proteins (Sørensen and Mortensen, 2005). The molecular chaperones DnaK and GroEL not only mediate proper de novo folding of proteins but are also involved in degradation of abnormal polypeptides by targeting them to proteases such as Lon and ClpP (Martínez-Alonso, et al., 2010). The holding chaperones IbpA and IbpB stabilize partly unfolded proteins without actively promoting their refolding and are often found tightly associated to the target proteins in inclusion bodies (Allen, et al., 1992). The disaggregating chaperone ClpB cooperates with DnaK and GroEL in reversing protein aggregation (Weibezahn, et al., 2004).

Strategies to improve specific productivity and solubility The goal in recombinant protein production is to obtain as much as possible of high quality product. A high total productivity can be achieved either by scaling up the cultivation volume, by increasing the biomass or by increasing the specific productivity (the productivity per cell). Promoter strength, inducer concentration, the plasmid copy number and the specific growth rate of the producing cells are some factors that have been shown to influence the specific productivity, but also the solubility of the product (Sandén, et al., 2005; Sandén, et al., 2002; Swartz, 2001; Tegel, et al., 2011; Terpe, 2006). When aiming at producing soluble proteins, the strategy is in general to slow down the synthesis rate of the protein. Examples of such strategies include using weaker promoters (Tegel, et al., 2011), using lower inducer concentrations (Sandén, et al., 2005) and to lower the specific growth rate of the producing cells (Sandén, et al., 2005). However, the opposite strategy, to use high synthesis rates during the production, increases the specific productivity (Ryan, et al., 1996; Sandén, et al., 2002). To find the best production conditions in order to obtain a large amount of soluble protein therefore relies on finding the correct balance between a high specific productivity and a high solubility.

17 The strength of a promoter is determined by the relative frequency of transcription initiation. This is affected by the affinity of the promoter sequence for the RNA- polymerase but also by the transcription rate of the RNA-polymerase. The T7 promoter, which is a very common promoter, requires host strains coding for the T7 RNA polymerase (Terpe, 2006). The E. coli BL21(DE3) strain and its derivatives, contain a lambda lysogen DE3 with the T7 RNA polymerase under the control of the IPTG inducible lacUV5 promoter (Terpe, 2006). The T7RNA polymerase transcribes RNA five times faster than the E. coli RNA-polymerase (Golomb and Chamberlin, 1974). In many cases the use of the T7 leads to high product accumulation, in some cases as high as 40-50% of the total cell protein (Studier and Moffatt, 1986). Although the high product accumulation is desirable, the use of strong promoters such as T7, has its drawbacks. High-level overexpression of genes can cause ribosome destruction (Dong, et al., 1995) and cell death (Miroux and Walker, 1996) and the use of strong promoter systems often results in the inability of the cell to fold the target protein properly (Baneyx, 1999). Further, production of mRNA is energy demanding and increases the metabolic burden on the cell.

Tegel and co-workers (2010) studied the effect of promoter strength on overexpression of proteins during batch cultivation in shake flasks. The total production as well as the soluble fraction of protein epitope signature tags (PrESTs), short regions of human proteins, was measured. The expression from the T7, the lacUV5 and the trc promoter was compared. In general, production under the control of the T7 promoter resulted in the largest total amount of target protein, whereas the use of the lacUV5 promoter, the weakest promoter in the study, resulted in the lowest amount of total protein. The weakest promoter generated the largest fraction of soluble protein and vice versa, and generally, the fraction of soluble protein was small using both T7 and trc. However, the amount of soluble protein was highest using T7 even though the soluble fraction was the lowest. This is explained by the much higher total production using this promoter (Tegel, et al., 2011).

18 The production of recombinant β-galactosidase was studied with respect to the specific growth rate at the time-point of induction (Sandén, et al., 2002). Induction was performed at specific growth rates of 0.5 h-1 and 0.1 h-1, respectively. It was shown that induction at the higher growth rate resulted in an almost 100 % higher specific productivity. The amount of mRNA formed was the same at both occasions and was therefore not the limiting factor. The ribosomal content, represented by the rRNA amount, was five times lower at the low specific growth rate. At the high specific growth rate, degradation of the ribosomes after induction, as a consequence of the high product formation rate, was also observed. In the study, the conclusion was drawn that translation, and not transcription, was the limiting factor in the protein synthesis capacity. Ryan et al (1996) also showed that cells growing at a high specific growth rate at the time of induction were more efficient in synthesizing the recombinant protein. The synthesis rate also remained higher for fast growing cells (Ryan, et al., 1996).

The concentration of IPTG and the feed rate of glucose during fed-batch cultivation was shown to influence the quality of the recombinant product when production was controlled from the lacUV5 promoter (Sandén, et al., 2005). The model proteins in the study were maltose binding protein (MBP) and a mutated form of this protein that was less stable and more prone to aggregation. Two different exponential feed profiles were used which resulted in constant specific growth rates of 0.2 h-1 (low feed rate) and 0.5 h-1 (high feed rate), respectively. Both the soluble and the insoluble fractions of the proteins increased with the inducer concentration. Using the highest concentration of inducer did not result in more soluble protein but in increased formation of inclusion bodies. The lowest amount of soluble protein was achieved by a high feed rate with the mutated, unstable form of the protein, in combination with a high inducer concentration. A high feed rate in combination with a high inducer concentration leads to a high syntheses rate of the protein. In this situation, the folding machinery of the cell might get overloaded and is unable to stabilise and fold the protein properly, which leads to increased proteolysis and aggregation.

Another factor that influences the efficiency of translation is the translation initiation region (TIR) of the mRNA. The main elements of translation initiation in E. coli are

19 the Shine-Dalgarno sequence, the initiation codon that in E. coli is mostly AUG, and a downstream region (DR). The SD is located 5-9 bases upstream of the initiation codon and is important for the binding of the mRNA to 30S ribosomal subunit. The sequence of the SD, in comparison to consensus sequence in E. coli, and its distance to the initiation codon are factors that determines the translational efficiency. The down stream region (DR) located immediately after the initiation codon also influences the gene expression level (Stenström and Isaksson, 2002). Engineering the translation initiation region of the mRNA is a promoter independent strategy of influencing gene expression. However, manipulations in the DR will in many cases change the amino acid sequence of the product (Baneyx, 1999).

A possible strategy for increasing the solubility of the recombinant proteins is by co- expression of chaperones. There are several chaperones or chaperone sets that have been selected for over production along with the recombinant target protein. The beneficial effect of co-expression of chaperones is however unpredictable and is a matter of trial and error, and there is no guarantee that the overexpression of the chaperones improves the solubility of the recombinant product (Sorensen and Mortensen, 2005). Co-expression of the chaperones DnaK and GroEL/ES might instead trigger proteolytic activities in the cell, which results in reduced yield, stability and quality of the recombinant protein (Martínez-Alonso, et al., 2010). The dual role of the chaperones, acting both as folding modulators and as proteolytic enhancers might partly explain the diverse results reported from co-expression studies (Martínez-Alonso, et al., 2010). Moreover, overproduction of chaperones contributes to the metabolic burden of the cell, which might explain the reduced yield of the recombinant protein.

Examples of other strategies that can be used in order to increase the solubility include cultivation at low temperature and the use of fusion tags (Sørensen and Mortensen, 2005).

20 1.4.2 Periplasmic production

The transport to the periplasm The main protein-translocation machinery in the inner membrane of E. coli is the Sec- translocase, which transport proteins across or insert proteins in the inner membrane. The substrates for the Sec-translocase all contain hydrophobic N-terminal regions. Proteins that are destined for secretion have N-terminal signal sequences that are processed by signal peptidase that removes the signal sequence after the translocation and allows release and folding of the protein in the periplasm. Inner membrane proteins have membrane-anchoring signals in their N-termini that most often remain associated with the inserted protein. The Sec-translocon can facilitate transport across or integration of proteins into the membrane in a co-translational or post-translational manner (Fig 6). The co-translational pathway is mainly employed for inner membrane proteins, while the secretory proteins mainly utilize the post-translational pathway. The selection of pathway lies at the early stage of translation when the nascent peptide emerges from the ribosomal exit tunnel. Very hydrophobic signals that emerge from the tunnel are bound by the signal recognition particle (SRP), which takes the complex of the ribosome and the nascent chain to the membrane-associated SRP- receptor FtsY. Elongation of the chain and insertion of the protein into the membrane via the translocon occurs simultaneously. The membrane protein YidC (not included in Fig 6) has also been shown to play a role in the insertion of a subset of membrane proteins via the translocon. If the signal sequence that emerges from the ribosomal exit tunnel does not display a high level of hydrophobicity, it is bound by the trigger factor (TF), which shields the nascent polypeptide from binding SRP. The newly synthesised protein is maintained in an unfolded state by the cytoplasmic chaperone SecB, which also targets the protein to the membrane bound ATPase SecA. Protein translocation across the membrane occurs at the translocon (du Plessis, et al., 2011).

Figure 6. Schematic and simplified representation of protein targeting to the Sec-translocase. i) The post-translational pathway: SecB binds to preproteins with less hydrophobic signal sequences and targets the preprotein to SecA which is associated to the translocase SecYEG. The signal sequence is cleaved off at the periplasmic side by the signal peptidase. ii) The co-translational pathway: More hydrophobic signals are bound by the signal recognition particle (SRP) as they exit the ribosomal tunnel. SRP takes the complex of the ribosome and the nascent chain to the membrane-associated SRP-receptor FtsY. Elongation of the chain and insertion of the protein into the membrane via the translocon occurs simultaneously.

Proteins may also be transported through the inner membrane by the twin arginine translocation (TAT) system. This system transports proteins having specific twin arginine motifs in their signal peptides and the proteins are transported in a folded state (Berks, et al., 2005).

Targeting recombinant proteins to the periplasm Recombinant proteins can be targeted to the periplasm by fusing natural signal sequences to their N-termini. Periplasmic expression is associated with several advantages in comparison with cytoplasmic production (Table 1). These include; (i)

22 authentic N-termini of the proteins are obtained after removal of the signal peptide by the leader peptidase, (ii) possible formation of disulphides in the periplasm due to the presence of the Dsb machinery (iii) lower amount of proteases in the periplasm compared to the cytoplasm (iv) simplified purification of the target protein if the periplasmic proteins are selectively released by osmotic chock or other strategies v) higher solubility of the product (Baneyx and Mujacic, 2004; Mergulhão, et al., 2005).

One difficulty with secretion is the inefficient export of the proteins across the inner membrane. This may result in degradation or in inclusion body formation or in the jamming of the membrane (Baneyx and Mujacic, 2004). Using signal sequences with increased hydrophobicity may direct the preprotein to SRP-pathway and thus eliminate toxic effects that might come from membrane jamming since the SRP- pathway has a tighter coupling between translation and translocation than the SecB- pathway (Wilson Bowers, et al., 2003).

The export capacity of the Sec-translocase was also studied by Mergulhão and Monteiro (2004). Different expression levels of two human proinsulin fusion proteins were accomplished by using different promoters and copy number plasmids. The periplasmic amount of product was almost the same although a 7 to 11-fold difference in the total expression level was obtained. The study shows that the translation level does not affect the maximum translocation efficiency and that a high synthesis rate of the product is only wasting the resources of the cell (Mergulhão and Monteiro, 2004).

Folding and degradation in the periplasm The periplasmic chaperones are involved in the folding of the periplasmic proteins and in the incorporation of proteins in the outer membrane. The Dsb proteins enable the formation and reshuffling of disulphide bridges. SurA and FkpA are examples of peptidyl-prolyl isomerases that are present in the periplasm (Moat, et al., 2002). Skp is and example of a general periplasmic chaperone that binds its substrate in a cavity, thereby protecting it from degradation (Walton, et al., 2009).

There are fewer proteases in the periplasm than in the cytoplasm. One example of a periplasmic protease is DegP. DegP function both as a chaperone and as a protease

23 and is essential for the removal of misfolded and aggregated inner membrane and periplasmic proteins. Synthesis of DegP is controlled by the σE regulon, which is activated in response to unfolded proteins in the periplasm (Missiakas, et al., 1996).

In order to increase the solubility of the product, the co-expression of chaperones may be efficient. In a study by Sonoda and co-workers (Sonoda, et al., 2011) this effect was studied on production of a single-chain Fv antibody secreted to the periplasm. It was found that the overexpression of the periplasmic chaperones Skp and of FkpA greatly increased the solubility of the product. However, co-expression of both FkpA and Skp had no synergistic effect. Further, co-expression of the cytoplasmic chaperones also affected the binding activity of the antibody fragment in the periplasm, especially the co-expression of DnaKJE.

Boström and co-workers (Boström, et al., 2005) studied the effect of the feed rate on recombinant protein secretion and degradation. Secretion of a protein to the periplasm resulted in less degradation and in the avoidance of the stringent response. It was also found that accumulation of acetic acid was ten times lower at a high specific growth rate when the product was secreted to the periplasm.

Leakage to the medium Periplasmic products show a tendency to leak to the culture medium, which is not desirable in most periplasmic processes. The tendency of leakage can however be utilized in processes where the recombinant proteins are targeted to the extracellular medium. There are a number of hypotheses in the literature concerning the E. coli cell´s inability to retain periplasmic products/leak periplasmic products to the medium (Shokri, et al., 2003). A common basis is that the structural integrity of the membrane, such as protein and lipid composition, is crucial for determining the retention. An altered composition of the membrane may result from genetic changes or from environmental changes such as medium composition or the feed rate of glucose (Shokri, et al., 2002).

24 Some E. coli mutants leak periplasmic proteins to the medium in a large extent. These mutants generally lack structural elements of the membranes or the cell wall, for example, LPS (Tamaki, et al., 1971) and murein lipoprotein (Lpp) (Nikaido, et al., 1977). Lazzaroni et al (1981) showed that mutants with a changed composition of outer membrane proteins (OMP), in this case a lower amount of OmpF, released more periplasmic protein to the medium (Lazzaroni and Portalier, 1981).

Secretion of overexpressed proteins to the periplasm also influences leakage. Cells that were secreting β-lactamase were more sensitive to detergents and they also had a higher non-specific release of periplasmic proteins to the medium. Analysis of the outer membrane protein composition showed that the amount of OmpA and OmpC was lower (40-60%) in secreting cells (Georgiou and Shuler, 1988).

Some signal peptides seem, for unknown reasons, to enhance export of periplasmic proteins to the medium. This was for example shown for insulin-like growth factor I that was secreted by the use of the signal peptide from staphylococcal protein A. This signal peptide has an extension in the N-terminus when compared to other signal peptides (Abrahmsén, et al., 1986).

1.4.3 Production in the inner membrane

Many membrane proteins can be overexpressed in inclusion bodies in the cytoplasm of E. coli, but their refolding into actively and functional proteins is difficult (Kiefer, 2003). The production thus relies on the expression in the cytoplasmic membrane from where the protein can be purified after detergent extraction (Wagner, et al., 2006). The accumulation of overexpressed proteins in the membrane is often toxic to the cell, which results in low yield and low biomass formation (Wagner, et al., 2006). The limited space in the cytoplasmic membrane, where the product can accumulate, is one potential bottleneck in membrane protein overexpression. The formation of a high biomass is thus of outermost importance for achieving high product yields (Wagner, et al., 2008).

25 Membrane proteins are usually targeted via the Sec-pathway to the Sec-translocon where they are inserted co-translationally into the membrane. Another potential bottleneck in membrane protein overexpression is the saturation of Sec-translocon (Essen, 2002). In a study by Wagner and co-workers (Wagner, et al., 2007), it was shown that membrane overexpression resulted in cytoplasmic aggregates containing the overexpressed proteins, chaperones and proteases. The aggregates also contained precursors of periplasmic and outer membrane origin, which indicates jamming of the Sec-translocon.

Co-expression of chaperones and cultivation at lower temperatures are common strategies that are used in order to facilitate folding and to reduce the level of inclusion body formation in membrane protein production. The CorA, the major magnesium transporter form the bacterial inner membrane, was used as a model protein (Chen, et al., 2003). In this study it was shown that lowering the cultivation temperature below 37°C reduced the levels of inclusion bodies formed and that a higher fraction of the CorA was incorporated in the membrane. The co-expression of the cytoplasmic chaperones DnaK/DnaJ and GroEL/GroES also increased the incorporation of CorA in the membrane. A conclusion from this study is that aggregation, and thus the formation of inclusion bodies, can be prevented by either reducing the protein synthesis rate or by increasing the cytosolic concentrations of DnaK/J and GroEL/ES.

The engineering or selection of host strains that are well suited for membrane protein overexpression is another possible approach for improving membrane protein overexpression. Two derivatives of BL21(DE3), named C41(DE3) and C43(DE3), were isolated for their ability to produce eukaryotic membrane proteins at elevated levels without toxic effects (Miroux and Walker, 1996). Through subsequent genomic and proteomic studies (Wagner, et al., 2008), it was shown that the original lacUV5 promoter of these strains, which is generally stronger than the original lac variant, had reverted to the original wild type promoter. It was shown that the lower strength of the lac promoter led to a lower transcription rate of the T7 RNA polymerase, and that this had a positive effect on the cellular viability leading to higher total production levels.

26 1.4.4 Surface display of proteins in E. coli

Proteins that are displayed at the surface of the E. coli cell have to be translocated through the inner membrane and pass the periplasm before they are inserted in the outer membrane. Before discussing surface display systems in E. coli, a short description of the outer membrane proteome and the insertion process of proteins in the outer membrane will therefore be given.

The outer membrane proteome Murein lipoprotein exists in a large number of copies in each cell and anchors the peptidoglycan layer to the outer membrane. OmpA, a monomeric protein that spans the outer membrane, is also important for the stability of the cells since it is cross- linked to the peptidoglycan layer (Moat, et al., 2002). The outer membrane contains many porins, trimeric proteins forming transmembrane β-barrel pores in the outer membrane that allow relatively unspecific passage of small hydrophilic molecules. OmpF and OmpC are some examples of porins. The porins can represent as much as 2 % of the total protein content in E. coli, making porins some of the most abundant proteins in terms of mass (Nikaido, 1996).

Maltoporin B (LamB) forms a channel in the outer membrane that specifically allows passage of maltose and maltodextrins (Nikaido, 1996). LamB also contributes to the ability of the bacteria to take up other sugars from the surroundings during glucose limiting conditions (Death, et al., 1993).

OmpT is a protease present in the outer membrane. It cleaves peptides between two consecutive basic amino acids (Kramer, et al., 2000). OmpT has been associated with the degradation of many recombinant proteins (Baneyx and Georgiou, 1990). It is stable under denaturating conditions and hence acts during purification and refolding processes (White, et al., 1995).

27 Insertion of proteins in the outer membrane A network of proteins is involved in the folding and insertion process of outer membrane proteins. The β-barrel assembly machinery (BAM) complex and a number of periplasmic chaperones, especially SurA, Skp and Deg P, play important roles in this process (Fig 7). The insertion process is ATP-independent due to the lack of a periplasmic ATP-pool (Knowles, et al., 2009).

SurA function both as a peptidyl-prolyl isomerase and as general chaperone. The main role of SurA in OMP maturation is that of a general chaperone (Sklar, et al., 2007). SurA has a high specificity for outer membrane proteins due to a specific pattern of aromatic residues that is found frequently amongst this group of proteins. SurA deficient mutants have a reduced amount of outer membrane proteins (Hennecke, et al., 2005).

The chaperone Skp has a selectivity for denatured OMPs (Chen and Henning, 1996). Skp deficient mutants have shown decreased levels of outer membranes proteins, for example decreased amounts of LamB, OmpA and OmpF. The skp gene is located downstream of BamA (one of the genes in the BAM complex) and both of these genes are regulated by the σE stress response that, for example, is activated in response to unfolded OMPs (Sklar, et al., 2007). Skp mediates transport of OmpA in an unfolded state across the periplasm (Walton, et al., 2009).

The last periplasmic chaperone that has a large impact on OMP biogenesis is DegP. This chaperone has both protease and general chaperone activity, and it was suggested that there are two different pathways for periplasmic chaperone activity in OMP maturation (Sklar, et al., 2007). SurA is the main chaperone during normal conditions whereas Skp/DegP have a primary role of rescuing OMPs that have fallen off the normal assembly pathway. However, under stress, or in the absence of SurA, the importance of Skp/DegP is increased.

Figure 7. A model for the periplasmic chaperone-mediated biogenesis of OMPs in E. coli. The OMPs are transported to the periplasm by the Sec-pathway. Most OMPs are escorted through the periplasm by SurA while off-pathway OMPs are rescued by the DegP/Skp complex. The DegP/Skp complex can deliver OMPs back to SurA, directly to the BAM-complex or use the protease activity of DegP for degradation. Modified from (Sklar et al. 2007).

β-barrel assembly machinery (BAM) complex The E. coli BAM-complex is composed of five domains; BamA, BamB, BamC, Bam D and BamE (Fig 7). The names of these proteins have recently changed and the former names were: BamA (YaeT, also called Omp85), BamB (YfgL), BamC (NlpB), BamD (YifO) and BamE (SmpA). BamA is an integral membrane protein and the other four proteins are lipoproteins that are localized to the inner leaflet of the outer membrane. All of the five BAM-proteins are involved in the OMP biogenesis but their specific roles in the process are not clear (Knowles, et al., 2009).

Surface display systems in E. coli The display of recombinant proteins at the surface of the bacteria is useful in many different biotechnological applications. It may be applied in the field of live-vaccine development, peptide library screening, whole-cell biocatalysis, biosensor development and bioadsorbents (Jose and Meyer, 2007). Surface expression of proteins in bacteria commonly relies on a fusion between the target protein (the passenger) and a carrier protein that usually is a cell surface protein naturally present in the host. For examples of surface display systems that have been used in E. coli,

29 see table 2. A successful carrier, (i) has an efficient signal peptide that transports the fusion protein through the inner membrane, (ii) has a strong anchoring structure so that the targets stays attached to the surface, (iii) is compatible with the target protein, meaning that stability of the carrier is not influenced by the target, and (iv) is resistant against degradation by proteases. The passenger protein itself also affects the translocation process. The folding structure of the passenger, such as the presence of disulphide bridges, is one factor that influences the success of the surface display. The size of the passenger, as well as the presence of certain amino acids in the passenger, are also known to influence the process (Lee, et al., 2003).

Table 2. Selected examples of surface display systems in E. coli. Advantages and disadvantages associated with each display system.

Display system Advantages Disadvantages

OMP Limited to transport of small passengers

Lpp´-OmpA Transport of large passengers Difficult to transport proteins with S-S Intermediate number of passenger bridges copies per cell

Surface organelles High number of passenger copies per Limited to transport of small cell passengers Difficult to transport proteins with S-S bridges

Autotransporters Transport of large passengers In some cases difficult to transport High number of passenger copies per proteins with S-S bridges cell Dimerisation at the surface

Fusions to OMPs and lipoprotein The first examples of recombinant surface display in E. coli was reported more than two decades ago and relies on fusions between outer membrane proteins such as OmpA, Lam B and OmpC and the passenger proteins. This type of display systems has been used for display of different peptides but it is not suitable for display of large proteins as the upper limit seems to lie between 60 and 70 amino acids (Charbit, et al., 1988).

30 The Lpp’OmpA hybrid display system is based on a fusion between the signal peptide and the first nine amino acids of the E. coli lipoprotein (Lpp) and three or five membrane spanning loops from the E. coli outer membrane protein A (OmpA). This system is not that sensitive to the sizes of the passengers, but appears instead to be sensitive to the presence of disulphide bridges in the passengers (Stathopoulos, et al., 1996). Intermediate numbers, approximately 104 recombinant passengers, has been reported to be expressed on the surface of the cell (Francisco, et al., 1993).

Fimbriae and other polymeric surface organelles Surface structures like fimbriae and flagella are present at high numbers at the surface of gram negative bacteria and are therefore attractive for display purposes. A large variety of fimbriae proteins have been used for surface display of different peptides. One drawback associated with this system is that there seems to be a restriction with regard to passenger size and composition. The successfully transported peptides are not larger than 10-30 amino acids and they are relatively hydrophilic. The formation of disulphide bridges in the passengers also seems to be a critical factor for transport (Klemm and Schembri, 2000).

Autotransporters Pathogenic gram-negative bacteria transport various kinds of virulence factors to the surface, or to the exterior of the cell, by the autotransporter pathway. The autotransporter protein family represents a large and highly diverse group of proteins that can be used as carriers for recombinant proteins. The passenger proteins that are naturally transported vary in length from 20 to 400 kDa and are also highly variable in sequence (Dautin and Bernstein, 2007). The autotransporters are synthesised as precursor proteins with a cleavable signal peptide, an N-terminal passenger domain and a C-terminal translocator domain composed of a α-helical linker region and a pore-forming β-barrel part (Fig 8). The signal peptide directs the preprotein to the Sec translocon and initiates the translocation process of the precursor protein across the inner membrane. The signal peptide is cleaved off and the β-domain becomes integrated in the OM. The α-helical linker is thought to form a hairpin structure in the pore and pull the passenger domain towards the surface (Jose, 2006). The autotransporters initially got their name because it was thought that they contained

31 everything that was needed for the transport and incorporation into the outer membrane in their own sequence. Today, there are several co-existing models describing how the passengers are actually transported to the surface of the cell. The model described in figure 8 is the hairpin model, which is the “classical” model. One of the newer models suggests the involvement of the BAM-complex in the incorporation process (Dautin and Bernstein, 2007).

There are several examples in the literature where different autotransporters have been used for surface expression of recombinant proteins in E. coli. The AIDA-I (adhesin involved in diffuse adherence) autotransporter from E. coli and the IgA protease from Neisseria gonorrhoeae are two of the most frequently applied autotransporters for this purpose (Jose, et al., 2002; Jose, et al., 2009; Jose and Meyer, 2007; Jose and von Schwichow, 2004; Klauser, et al., 1990; Li, et al., 2008; Maurer, et al., 1997).

A medium! MECHANISM FOR AUTOTRANSPORT!

OM!

periplasm!

IM! cytoplasm!

B Passenger! Translocation unit (TU)! SP! N! C!

#-barrel pore! "-helical linker region!

Figure 8. The autotransporter system. A) Suggested mechanism for autotransport. The signal peptide is cleaved off following translocation across the inner membrane through the Sec-translocon. The pore- forming part of the translocation unit forms a pore in the outer membrane and pulls the passenger towards the surface of the cell. B) Structure of an autotransporter vector. The vector includes a signal peptide (SP), a passenger protein and a translocation unit that is composed of a linker region and a pore-forming part.

32 1.4.5 Extracellular production

Recombinant proteins can also be targeted to the medium, which offers the benefits of low levels of proteolysis, simple detection and purification, a better folding environment and the avoidance of the N-terminal methionine extension. Extracellular production of toxic proteins minimizes their impact on the host and may represent the only possible way for their production (Hannig and Makrides, 1998; Ni and Chen, 2009). The high dilution of the product may however be disadvantageous. One strategy is to utilize leaky mutants that release periplasmic proteins to the medium. Another strategy relies on import of secretion mechanisms, either lysis or whole transport mechanisms, from pathogenic E. coli (Shokri, et al., 2003).

The haemolysin transport system is one candidate for transport of the recombinant products to the medium. The proteins are translocated directly to the medium in a protein channel that spans both of the membranes. The channel is composed of HlyB/HlyD and TolC and the signal peptide of HlyA is fused to the product and directs it to the protein channel. The proteins that form the channel also need to be co- expressed which is one disadvantage with the system. However, transport of the product by the haemolysin system does not compete with transport of endogenous proteins, which is advantageous (Blight and Holland, 1994).

Periplasmic products can also be released to the medium by import of lysis mechanisms. This strategy is based upon the action of colicin lysis proteins where the colicin E1 lysis protein (kil) has been mostly used. The induction of the lysis protein results in permeabilisation of the outer membrane and the subsequent release of products (Miksch, et al., 1997).

The autotransporters can also be utilized for extracellular transport of recombinant proteins since some autotransporters release their passengers to the medium (Jose and Meyer, 2007).

33 2 PRESENT INVESTIGATION

With the completion of the human genome project (HUGO) during recent years, there is today a large need for production of proteins, not only in order to determine their structures and functions, but also as they may constitute proteins of large value for the biopharmaceutical industry. However, since proteins are very diverse and different from each other, it is virtually impossible to know how to produce a certain protein in the most suitable way just by looking at the amino acid sequence. Thus, in order to increase the probability of finding production conditions appropriate for a particular protein, there is a need for an increased understanding of production strategies already available as well as the establishment of new and better production methods.

Proteins that are used for research purposes are preferably produced in an active, soluble form. Typically, the proteins are produced in small-scale cultivations using batch technology. Unfortunately, this technology suffers from drawbacks related to e.g., oxygen limitation, acetic acid production and low biomass, but is in spite of that a frequently used method since it is so easy to perform. However, cells grow at their maximal specific growth rate in a batch culture, which in many cases is not optimal when aiming at producing active soluble proteins, as described earlier.

In the biopharmaceutical industry, the fed-batch technology is commonly used for control of the glucose uptake rate. The glucose uptake rate, or the feed-profile, has in the literature been described to influence parameters such as specific productivity, solubility and proteolysis (Boström, et al., 2005; Ryan, et al., 1996; Sandén, et al., 2005; Sandén, et al., 2002). Some proteins cannot be produced in an active form in the cytoplasm and are therefore instead secreted to the periplasm, but this production strategy is usually associated with low yields. Other “difficult-to-express” proteins, e.g., proteins that are toxic to the cell, are also preferably transported to parts of the cell other than the cytoplasm. This will not only provide another possibility for successful protein production but also simplified purification as the number of endogenous proteins are significantly less in the periplasm or outside the cell than in the cytoplasm.

34 A compartment that has been neglected for long from the perspective of large-scale protein production is the cell surface. Despite this, bacterial surface display systems has during the last 20 years emerged as an important research tool, e.g., display of peptide/protein libraries in various protein engineering efforts, display of protein immunogens for the construction of live vaccines, for biocatalysis and bioremediation, recently reviewed in (Daugherty, 2007; van Bloois, et al., 2011). Potentially, bacterial display could be an attractive alternative to more established large-scale protein production strategies, as it offers a simple purification scheme and possibilities for on-cell protein characterisation. However, the utility of surface display in this context (large scale protein production) has to be proven by further investigations.

As an alternative to traditional fed-batch technique, we here describe an approach for control of the glucose uptake based on mutant bacterial strains deficient in the phosphotransferase system (PTS) (Paper I), and its impact on cell density, specific productivity, acetic acid formation and oxygen consumption (Paper I & III), Historically, there have been very few reports on the impact of the feed-rate on processes with product localisations other than the cytoplasm. Thus, we also investigated how the glucose feed-rate affects recombinant protein production in the periplasm (Paper II) and when displayed on the bacterial surface (Paper IV).

2.1 A cellular alternative to fed-batch cultures (I, III)

2.1.1 Strain evaluation (I)

The aim of this part of the study was to find a method to control the glucose uptake rate in small-scale batch cultivations where the traditional fed-batch technique is not applicable. The strategy was to use strains with mutations in the phosphotransferase system (PTS) that is involved in the glucose uptake. We reasoned that the mutants would have lower growth rates due to the mutations and as a result of that also lower oxygen consumption and decreased formation of acetic acid. Hopefully, this would result in the establishment of higher cell densities and improved product yield, and the strains would thus represent a novel, simple alternative to fed-batch cultures.

35 The strain AF1000, that originates from the E. coli K12 strain MC4100 (Casadaban, 1976; Sandén, et al., 2002) was used in this study and is referred to as wild type (WT) strain. This strain lacks all constituents of the lactose operon and it grows on minimal salt medium without any additional supplements. The three variants of this strain, with mutations in genes coding for proteins belonging to the PTS system (Picon, et al., 2005), that where used are presented in table 3.

Table 3. Strains with mutations in PTS. The mutations in the different strains are shown as well as the absent enzymes.

Original name Name of the strain Mutation in Absent enzyme of strains in this work gene cluster

Man PPA 668 PTSMan manX Enzyme IIAB

Glc PPA 652 PTSGlc ptsG Enzyme IICB

Man PPA 689 PTSGlcMan manX and Enzyme IIAB ptsG Enzyme IICBGlc

The E. coli homologous enzyme β-galactosidase, which hydrolyses lactose into monosaccharides, was used as a model protein since it is a stable protein and has a high solubility even at high concentrations. Another advantage with the model protein is the straightforward analysis for enzyme activity. The lacUV5 promoter, which is theoretically not subjected to catabolite repression (Arditti, et al., 1973), controlled the expression of the recombinant protein from a low copy number plasmid (pACYC). AF1000 and its derivatives are lac negative and production from the lacUV5 is thus constitutive. A system that could be induced by IPTG addition was also constructed by insertion of the F´-factor (lacIq, lacY, lacA).

Cell-growth First, we studied impact of the mutations on the growth rate of the mutant strains. All strains were therefore grown in batch cultivations with unlimited access to glucose. The cell density increased exponentially and the resulting growth rates were 0.78 h-1,

-1 -1 0.38 h and 0.25 h for PTSMan, PTSGlc and PTSGlcMan, respectively (Fig 9). This is to be compared to the wild type, which had a specific growth rate of 0.72 h-1. This growth rate can however not be considered as deviating from the one obtained for

36 PTSMan. This result was encouraging as mutations in the PTS system in a previous publication had been shown to reduce the glucose uptake rate (Picon, et al., 2005).

! = 0.72 5 ! = 0.78

! = 0.38 )

-1 4

! = 0.25 Cell dry weight (g l 2

0 0 2 4 6 8 10 12 14 16 Cultivation time (h)

Figure 9. Biomass accumulation as a function of cultivation time. Comparison of the wild type cell (WT) and the mutants grown in batch cultivation. Circles:WT, squares: PTSMan, diamonds: PTSGlc, triangles: PTSGlcMan.

Acetic acid and specific oxygen consumption rate

The next variable studied was the acetic acid formation. The wild type and PTSMan produced acetic acid in proportion to the growth rate (Fig 10). The PTSGlc and

PTSGlcMan produced small amounts of acetic acid. The lower production of acetic acid in PTS-mutants has been observed previously (Chou, et al., 1994).

A further step was to compare the mutants grown in batch with the wild type grown in fed-batch. Exponential feed rates were therefore designed to theoretically give the same growth rates for the wild type in fed-batch as the mutants had in batch. The

PTSMan had a specific growth rate approximately equal to that of the wild type, making this type of comparison irrelevant to perform for this mutant.

The mutants had a specific oxygen consumption rate that was proportional to the growth rate (data not shown). The oxygen consumption rate for the PTSGlcMan cultivated in batch was almost identical to the WT cultivated in fed-batch. This comparison was important to make in order to exclude the possibility that the

37 mutations resulted in increased respiration, since this would have had negative impacts on the establishment of high cell densities.

The acetic acid accumulation was close to zero in the fed-batch cultivations with the wild type as was it for the mutants in batch (data not shown). Batch cultivated mutants are thus equal to fed-batch cultivated WT with respect to acetic acid formation and oxygen consumption.

1200

1000

800 ) -1 600 HAc (mg L 400

200

0 0 2 4 6 8 10 12 14 16 Cultivation time (h)

Figure 10. Accumulation of acetic acid as a function of cultivation time. Comparison of the wild type cell (WT) and the mutants grown in batch

cultivation. Circles: WT, squares: PTSMan, diamonds: PTSGlc, triangles: PTSGlcMan.

Product formation The next parameter to evaluate was the specific product formation rate. The constitutive system was initially used for production of the model protein β- galactosidase. Figure 11 shows data from batch cultivations with the wild type and the mutants. The PTSMan strain had the highest and most stable specific production rate. The wild type also had a relatively high specific production rate in the beginning of the cultivation, but it dropped to around 50% at the time of harvest. PTSGlc had an initial low specific production rate that dropped even more as a function of time. The

PTSGlcMan had an even lower specific productivity (data not shown). By using cells containing an F´-factor with lacIq, the production could be induced by addition of

38 IPTG. The production was increased in the PTSGlc (as can be seen in figure 12) but remained low in PTSGlcMan. The reason to why the production is so low in the double mutant, PTSGlcMan, is not known. The low production might be a consequence of the low synthesis rate of proteins in this mutant. It is also possible that the absence of the two PTS enzymes, which are central for the glucose uptake in the cell, somehow affects the transcription from the substrate-induced promoter lacUV5.

50 ) -1 40 h -1

(U mg 30 p q

0 5 6 7 8 9 10 11 12 13

Cultivation time (h)

Figure 11. Specific production rate of β-galactosidase during batch cultivation of the WT strain and the mutants as a function of the

cultivation time. Circles: WT, squares: PTSMan, diamonds: PTSGlc.

The last step of this evaluation was to compare the specific production rate of the

PTSGlc cultivated in batch with the WT cultivated in fed-batch (Fig 12). The production was induced by addition of IPTG when the cells had a specific growth rate of approximately 0.4 h-1. The specific production rate was approximately 50 U mg-1 h-1 after the induction in both strains. A reduction in the specific productivity was seen at an earlier time-point for the PTSGlc than for the wild type. The total β-galactosidase accumulation at the end of the cultivation was estimated to be 33% and 13% of the total protein, for the wild type and PTSGlc, respectively.

39 30 60

50 25

) 20 -1 qp (U mg 30 15 -1 h

20 -1 )

Cell dry weight (g l 10 10

5 0

0 -10 -4 -2 0 2 4 6 8 10 12

Time from induction (h)

Figure 12. Dry weight and specific production rate of β-galactosidase of the PTSGlc mutant grown in batch and the WT strain grown in fed-batch at the corresponding growth rate. Circles: WT (fed-batch), diamonds: PTSGlc (batch). Closed symbols: cell dry weight, open symbols: specific production rate.

High cell density cultivation As a reduction of the growth rate, by the use of the mutants, potentially would result in higher cell densities in batch cultivations, we studied this. Therefore, the WT and the PTSGlcMan were grown in a repetitive batch mode where new batches of glucose were added as the glucose in the medium was consumed. This glucose addition strategy was chosen in order to avoid negative effects from high glucose concentrations (Lara, et al., 2008; Shiloach and Fass, 2005). The final cell mass

-1 -1 reached was 27 g l for the wild type and 34 g l for PTSGlcMan respectively (Fig 13). The WT produced ten times more acetic acid during the cultivation than the mutant,

-1 -1 9000 mg l compared to 900 mg l . It was noted that already at acetate levels of 500 mg l-1, the growth rate of the WT was affected, and at this time the cell concentration was only 3 g l-1. The shaded area in figure 13A shows the part of the cultivation that is principally not operational due to the high concentration of acetic acid. The PTSGlcMan (Fig 13 B) has constant but lower growth rate, a much longer period. At the end of the cultivation an increase in the acetic acid accumulation is observed. This acetic acid probably originates from mixed acid fermentation, which is a result of insufficient supply of oxygen to the fermentor.

1 0,4 A B 8000 8000

0,8 ! 0,3 !

6000 6000 ) )

-1 HAc -1 h h HAc (mg l 0,6 HAc (mg l

-1 0,2 -1 (g g (g g O2 O2 ), q ), q -1 4000 -1 -1

-1 4000 ) q ) 0,4 0,1 O2 my (h my (h

q O2 2000 2000 0,2 0

HAc 0 0 -0,1 0 0 2 4 6 8 10 12 0 5 10 15 20 25 Cultivation time (h) Cultivation time (h)

Figure 13. High cell density cultivations of A wild type (AF1000) and B PTSGlcMan mutant strain. Closed circles: specific growth rate which is also approximated by a solid straight line, open circles:

acetic acid concentration, solid line: specific oxygen consumption rate, qO2. The shaded area in figure A indicates the interval of cultivation that is preferably avoided.

2.1.2 Production of integral membrane proteins by the PTS-mutants (III)

A following logical step was to explore the possibility of using the mutants for production of other proteins, especially proteins that are known to be difficult to produce. Integral membrane proteins were chosen as model proteins since their production is usually associated with low yield, toxicity and inclusion body formation (Essen, 2002). A set of five different E. coli integral membrane-spanning proteins (IMPs) with human homologues were selected as model proteins (Table 4) and were produced with N-terminal His6 tags to facilitate purification. These proteins have earlier been expressed using the T7 promoter with varying success in different strains (Eshagi, et al., 2005), but in the present study the expression was controlled from the araBAD (PBAD) promoter. The B-strain, BL21(DE3)(Novagen) that is commonly used for production of membrane proteins was also included in the study. This strain has a specific growth rate of 0,76 h-1 in minimal medium, which is approximately equal to the specific growth rate of AF1000 (WT).

Table 4. Membrane protein characteristics. The integral membrane proteins used in the study.

Target Function/Predicted function

EM03 Ammonium transporter

EM09 Predicted transporter

EM16 Chloride transporter

EM20 Multi drug efflux system protein

EM29 Intramembrane serine protease

Small-scale production All strains that were included in the study were grown in minimal medium in micro titre plates, and expression of the IMPs was induced at OD600 ≈ 1 by addition of L- arabinose. All cells were harvested after the same time of expression. A dot blot (Fig

14) was used to show the expression of the targets. PTSGlcMan and PTSGlc produced three out of the five proteins, as did the BL21(DE3) strain. The amounts of cells at the time of harvest varied since the strains have different growth rates. A quantitative comparison of the expression levels in the different strains is not therefore possible from this dot blot.

!"#$ %!"#& %!"'( %!")# %!")&%

*+'###%

,-."/0%

,-.123%

,-.123"/0%

45)'67!$8%

Figure 14. Expression of membrane proteins. Dot blot of purified His-tagged proteins overexpressed in the following

strains: AF1000, PTSMan, PTSGlc, PTSGlcMan and BL21 (DE3).

42 Medium scale production and quantification In order to be able to compare the relative amounts of IMP´s and to assess the degree of homogeneity of the target protein, the production of the EM03 target was scaled up into shake flasks. The membrane fraction was subjected to IMAC resin in order to purify the His6-tagged product. Expression of the target protein was verified by Comassie-stained SDS-PAGE (Fig 15).

+,# !#####'#####(#####%######*#####&###### -./01# !""# $"#

&'# %$# ("#

'"#

!)#

Figure 15. Verification of expression of EM03. SDS-PAGE analysis of IMAC-

purified target EM03. 1 marker, 2 AF1000, 3 PTSMan, 4 PTSGlc, 5 PTSGlcMan, 6 BL21 (DE3).

The samples were further subjected to analytical gel filtration to allow a quantitative comparison of the expressed levels in the different strains. The gel filtration chromatogram is shown in figure 16A where values for the same amounts of cells are shown normalized so that the top peak (maximum) value equals 1. The chromatogram shows that EM03 could be purified as a non-aggregate.

43 Figure 16. Purification and quantification of the selected membrane protein. (A) Analytical gel filtration from medium-scale shake flask cultivations using the target protein, EM03. The

chromatogram shows values for an equal OD6oo -value of cells in the different samples, normalized with respect to the highest top peak absorbance set equal to one. (B) Estimated peak areas of analytical gel filtration chromatograms from the production of the target EM03. The values are

shown as calculated peak areas for an equal OD600 -value of cells in the different samples, normalized with respect to the largest top peak area set equal to one.

The amounts of EM03 that were produced in the different strains were calculated from the areas under the curves and the results are plotted in figure 16B. The comparison shows that PTSGlc produced approximately twice as much protein as

PPA689 PTSGlcMan and BL21(DE3). AF1000 and PTSMan produced very small

44 quantities of EM03 also in shake flasks, so that it was not possible to determine the areas in a statistically relevant fashion.

It is obvious that the glucose uptake rate influences the amount of protein that can be extracted in a soluble form from the membrane, but this is apparently not the only factor since AF1000 and BL21 have the same growth rates but very different production levels.

Acetic acid As stated before, the by-product acetate has a negative impact on both the growth rate and on the production of recombinant proteins (Eiterman and Altman, 2006; Jensen and Carlsen, 1990; Koh, et al., 1992). It is difficult to separate the effect of the acetic acid from the effect of the growth rate on the production, since a growth rate above a certain threshold value leads to the excretion of acetic acid. In some studies, acetic acid has therefore been added to the cultivation and its effect on growth rate and on production of recombinant proteins has been studied. Jensen and Carlsen (1990) showed that acetic acid concentrations of above 6.1 g l-1 resulted in decreased growth rate but in another study (Nakano, et al., 1997) this effect was seen already at concentrations of 500 mg l-1. In our study (I) AF1000 showed a reduction in growth rate at concentrations of 500 mg l-1 and at this time the cell concentration was approximately 3 g l-1.

Table 5. Yield of acetic acid per cell for the strains used in the study.

Strain HAc (mg/L) per OD600-unit

AF1000 158

PTSMan 147

PTSGlc 3

PTSGlcMan 2

BL21(DE3) 58

45 The BL21(DE3) strain has a high specific growth rate (approximately 0,76 h-1) which is comparable to the specific growth rate of AF1000, but has a comparably low production of acetic acid (Table 5). The production of the membrane protein target

EM03 is twice as high in PTSGlc as in PTSGlcMan and BL21(DE3) and very low in

AF1000 and PTSMan. The differences in the accumulation of acetic acid between the strains might be one possible explanation for the different production levels.

However, this in not the only explanation since the PTSGlc and the PTSGlcMan has the same production of acetic acid, but different levels of production of EM03. The production level is probably a combination of the glucose uptake rate and the acetic acid accumulation level. The BL21(DE3) strain is deficient in the proteases Lon and OmpT, which might also explain the relatively high production level in this strain.

2.2 Impact of feed-rate on processes with other product localisations than the cytoplasm (II, IV)

2.2.1 Leakage of periplasmic products in relation to the glucose uptake rate (II)

Some proteins need to be transported to the periplasm in order to fold properly. Leakage of periplasmic proteins to the medium is a problem associated with periplasmic production provided that the product ending up in the medium is not collected. The aim of the present study was to investigate how the glucose uptake rate is correlated to the leakage/periplasmic retention of secreted products. The growth rate was varied and its effect on leakage of two periplasmic products to the medium was studied. The focus was also on how the amount and specificity of the outer membrane proteins was affected by the growth rate and by recombinant protein production. Could the amounts of these proteins be connected to leakage of periplasmic products to the medium?

Cutinase, a lipolytic enzyme, from Fusarium solani pisi was used as one of the model proteins (Mannesse, et al., 1995). Two Z domains, i.e., engineered forms of the B domain of staphylococcal protein A, which is present in the cell wall of Staphylococcus aureus, were fused to the N-terminus of the cutinase (Bandmann, et

46 al., 2000). The expression of the product was controlled by the constitutive staphylococcal protein A promoter (Löfdahl, et al., 1983), and the signal peptide from the same protein was used to direct the recombinant product to the periplasm. An antibody fragment (Fab) was used as a second model protein, which was directed to the periplasm by an OmpA signal sequence. The production of Fab was under control of the tac promoter. ZZ-cutinase was expressed in the K12 strain 0:17 (Olsson and Isaksson, 1979) and Fab in W3110 (ATCC 27325).

Specific productivity and leakage The specific productivity of the model proteins followed the growth rate of the cells and was higher at higher growth rates. This trend was also observed for the production of β-galactosidase (I) and is in accordance with earlier findings (Sandén, et al., 2002). Figure 17 shows the leakage of ZZ-cutinase, the quotient between product activity in the medium and product activity in total, as a function of the growth rate. A higher specific growth rate resulted in a higher leakage of periplasmic product to the medium. The same trend was observed for the production of Fab (data not shown). Leakage is defined as the selective passage of proteins through the outer membrane, whilst maintaining a functional cell. This is distinguished from lysis, which leads to disruption and cell-death, but that also results in release of periplasmic products to the medium. The presence of protein in the medium corresponding to the normal endogenous composition of E. coli proteins (here referred to as total protein) is usually considered as an indication of lysis. Product activity and total protein in the medium was measured in order to exclude that the product found in the medium was a result of lysis. Not more than 4% of the total protein was found in the medium in the batch cultivation with recombinant product formation and not more than 3% in all other cultivations (data not shown). The leakage of the product to the medium was as high as 20% in the batch cultivation, and 9 and 6% respectively in the fed-batch cultivations. The main part of the ZZ-cutinase activity found in the medium was thus a result of leakage and not of lysis.

0.20

0.15

µ=0.3 h-1 (batch)

0.10

µ=0.2 h-1 (exp. fedbatch)

Cutinase activity (medium/total) 0.05 µ=0.1 h-1 (exp. fedbatch)

0 0 5 10 15 20 DW (g l-1)

Figure 17. Leakage of ZZ-cutinase to the medium in percent of the total production. The specific growth rates (h-1) are indicated in the figure.

Effects of glucose uptake rate and recombinant protein production on outer membrane composition In order to investigate the relationship between the content of outer membrane proteins and leakage, the first step was to identify which proteins that were present in the outer membranes during the cultivations. The outer membrane proteins were therefore isolated from cultivations with different growth rates (both with and without production of ZZ-cutinase). The most abundant proteins were identified as OmpA, OmpF and LamB by N-terminal sequencing (Fig 18). The expression of the outer membrane proteins involved in uptake of nutrients, LamB and OmpF, was dependent on the specific growth rate. The titre of these proteins decreased as the growth rate increased for both ZZ-cutinase and non-producing cells. However, the reduction was more noticeable for the producing cells. No LamB was detected in the batch cultivation, which was to some degree expected since Death and co-workers (Death, et al., 1993) showed that the LamB protein is upregulated during glucose limitation. The OmpA protein, which is thought to have a role in stabilizing the cell, seemed to be more or less constant regardless of the growth rate. There was however a slight tendency that the amount of the protein was lower in producing cells than in non- producing cells at comparable specific growth rates.

48 600 A

500

) 400 -1

300

OmpF (mg l 200

100

200 B 180

160

140 )

-1 120

100

80 LamB (mg l 60

20 not not det det 0

400 C 350

300

) 250 -1

200

150 OmpA (mg l 100

1000 D 900

800 Figure 18. Outer membrane proteins isolated from cultivations with different 700 growth rates (exponential feed profiles) ) -1 600 without production (black bars) and during 500 ZZ-cutinase production (white bars). Lam B was not detected in all cultivations, this is 400 OMP (mg l indicated by “not det” in the figure. 300 A. OmpF (mg l-1) 200 B. LamB (mg l-1 ) -1 100 C. OmpA (mg l ) D. Sum of OmpF, LamB and OmpA (mg l-1) 0 0,1 0,2 0,3 0,45 0,6

Specific growth rate (h-1 ) !

49 Figure 18D summarises the combined titre of proteins investigated. The combined titre of outer membrane protein decreased as the specific growth rate increased for both producing and non-producing cells. The overall titre of outer membrane proteins was lower if a recombinant protein was produced. The total reduction was almost 60% with a high product formation rate at µ= 0.3 h-1.

It is reasonable to hypothesise that outer membrane protein accumulation influences product retention. When studying the total titre of outer membrane proteins it was clear that the titre was inversely proportional to the growth rate and that a lower titre was found in producing cells. Secretion of a recombinant product results in an increased transport of proteins through the inner membrane. Our model system showed a specific productivity that was proportional to the specific growth rate, allowing different levels of transport through the Sec translocase to be studied. Depletion of outer membrane proteins, due to jamming and overloading of the sec system, has earlier been proposed as possible reasons for periplasmic leakage (Baneyx, 1999). In the present study, a higher growth rate, which is connected to a higher transport through the Sec translocon, resulted in a higher leakage and in a lower amount of outer membrane proteins. Inhibited transport of outer membrane proteins due to the overloading of the Sec translocon is therefore a possible explanation to our results.

Another approach that can be used when studying consequences of limited transport through the Sec translocon is by altering the levels of the components of the Sec translocon itself. Baars and co-workers (2008) used this approach and compared the outer membrane proteome of cells depleted of SecE with the outer membrane proteome of cells expressing normal levels of this protein. It was shown that the amount of most outer membrane proteins was reduced upon SecE depletion. However, the amount of OmpA was however unaffected, which is in accordance with our findings (Baars, et al., 2008).

In a previous study (Shokri, et al., 2002), it was shown that the growth rate influenced the composition of phospholipids and fatty acids in the membranes. The leakage of the periplasmic product, in that study β-lactamase, showed a peak at a dilution rate

50 corresponding to a specific growth rate of 0.3 h-1. At this specific growth rate, there was an increased amount of phosphatidyl glycerol and unsaturated fatty acids in the membrane. The expression level of the product in that study was low and maybe not enough for influencing the outer membrane composition. The growth rate intervals investigated in these two studies were different and therefore a direct comparison was difficult to do.

In the present study, we investigated how the composition of the outer membrane proteins was connected to leakage of recombinant proteins. Our results suggest that the permeability of the outer membrane is a function of both the lipid composition and its protein content.

300

250 ) -1

200

150

100 Cutinase activity in cell pellet (U ml 50

0 0 5 10 15 20 DW (g l-1)

Figure 19. Periplasmic retention of ZZ-cutinase (cell pellet) as a function of dry weight (DW). The specific growth rates are indicated in the figure as follows: a specific growth rate of 0.3 h-1 (filled circles), a specific growth rate of 0.2 h-1 (filled triangles) and a specific growth rate of 0.1 h-1 (open squares), respectively.

Figure 19 shows the retention of ZZ-cutinase, i.e. the amount of product that can be harvested from the cell paste, as a function of cell accumulation. Interestingly, the amount of product that can be retained within the cell is always the same and not a function of the growth rate. This means that a higher specific growth rate, which results in higher synthesis rate, will not result in more product in the periplasm of each cell due to the elevated leakage.

51 2.2.2 Optimisation of surface expression using the AIDA autotransporter (IV)

Although bacterial surface display has been used extensively in various research applications, there are few studies focusing on its use as a large-scale protein production tool. If surface-display technology is to be used in the context of large- scale production, methods for cultivation of large quantities of cells whilst maintaining a high surface expression is needed. Here, we investigated the influence of different cultivation conditions and techniques, batch and fed-batch, as well as the impact of two different cultivation media on the production of cell surface-displayed recombinant proteins.

As a model protein, we used the staphylococcal protein A-derivative Z, which was chosen because it is a small protein without disulphide bridges, and should theoretically not be difficult to transport to the surface. Further, Z binds to the Fc region of IgG, thus enabling flow cytometry-based verification and quantification of surface expression of Z by using fluorescent IgGs. The vector that was used for the surface display is based on the AIDA autotransporter with its own wild type promoter aidA (Fig 20).

C PaidA SP Z L AIDA

Figure 20. The AIDA-vector containing the aidA promoter, the signal peptide (SP), Z (the passenger protein), a linker (L) and the β-barrel pore (AIDAc).

Impact of OmpT on surface expression level of Z Z contains a potential cleavage site for the outer membrane protease OmpT. A first step of this work was to assess if the potential cleavage site was accessible for OmpT. Analysis of the AIDA-Z fusion protein expressed in strains with and without OmpT (0:17 (Olsson and Isaksson, 1979) and 0:17ΔOmpT, respectively) revealed that both strains expressed the fusion but only the OmpT negative strain yielded the full-length

52 protein (Fig21). The degradation product in 0:17 has approximately the same size as AIDAc, a size that is expected if OmpT cleaves within the Z-domain.

Figure 21. Comparison of surface expression of Z in OmpT positive and OmpT negative E. coli 0:17. (A) Western blot of the isolated outer membrane fraction, using antiserum against AIDAc. Lane 1: Marker, 2: 0:17, 3: 0:17ΔOmpT, 4: 0:17 with empty vector, 5: 0:17ΔOmpT with empty vector, 6: 0:17 with vector containing Z, 7: 0:17ΔOmpT with vector containing Z. (B). Cells analysed by flow cytometry using human IgG linked to AlexaFluor488. Red: 0:17 with vector without passenger, Blue: 0:17 with vector containing Z, Green: 0:17ΔOmpT with vector containing Z.

Flow cytometry analysis of the different clones confirmed that Z was accessible and functional at the cell surface, since binding to fluorophore-labeled IgG was achieved. The fluorescence intensity was, as expected from the western blot analysis, lower in 0:17 than in 0:17ΔOmpT. The strain 0:17ΔOmpT was thus chosen for the further experiments. The effect of the temperature on the expression level was also investigated and expression was found to be favoured by cultivation of the cells at 37°C.

Effects of growth medium and regulation of the aidA promoter In order to evaluate the effect of the growth medium on the surface expression of Z the cells were grown either in minimal medium with glucose as carbon source or in Luria Broth (LB) medium. As can be seen in figure 22, the expression level was higher in the minimal medium than in LB, which is in accordance with earlier findings (Benz, et al., 2010). Glucose was added to the LB medium in order to assess

53 if the lower expression level was a result of the absence of an easily available carbon source. This was not the case since the addition of glucose to the LB medium resulted in even lower expression levels (Fig 22). The concentration of amino acids in the medium was followed during the cultivation (data not shown). Some amino acids were depleted but their depletion did not result in increased expression levels of Z. Another idea that we had was that glucose starvation and the subsequent entrance into the stationary phase would affect the level of expression. The cells were thus grown in minimal salt medium containing glucose and were allowed to grow until the glucose was consumed. It was shown that entry into the stationary phase reduced the surface expression level of Z (data not shown). These data together with those from the cultivations in different media indicate that the aidA promoter is not induced by stringent response.

Figure 22. Comparison of surface expression of Z in 0:17ΔOmpT grown in glucose minimal medium (grey), LB medium (white) and LB medium with 10 g l-1 glucose (black), respectively.

Nevertheless, cultivation in different media does result in different growth rates. Going from minimal medium to LB medium and finally to LB medium supplemented with glucose result in progressively higher growth rates. Since it appeared likely that lower growth rates resulted in enhanced surface expression, the next logical step was to investigate if further reductions of the growth rate would be beneficial.

54 Effects of glucose uptake rate in minimal salt medium Different fed-batch cultivations were performed in order to investigate the influence of the glucose uptake rate/growth rate on expression. Three different growth rates (μ=0.1, 0.2 and 0.4 h-1) were achieved by exponentially feeding glucose to the reactors after an initial batch phase. Surface expression of Z was followed using flow cytometry. There is a tendency that the expression level is higher at a higher growth rate (Fig 23). The limited transport capacity of the Sec-translocon (Baars, et al., 2008; Baneyx, 1999; Mergulhão and Monteiro, 2004) is a potential bottleneck in production systems where the recombinant protein is localised to parts of the cell other than the cytoplasm. Likewise, the insertion of proteins in the outer membrane, mediated by the BAM-complex, is a potential bottleneck in surface display systems. The outer membrane protein fraction was isolated from the cultivations with different growth rates (data not shown) and no clear difference in outer membrane protein content could be seen. In (II) those differences appeared clear. The expression level of Z by the aidA promoter is probably not high enough for influencing transport and insertion of endogenous proteins and the pattern of protein expression therefore probably reflects the synthesis rate of the protein.

Fig 23. Cell growth and antibody-binding activity of the surface displayed Z. OD600 values (filled symbols) from the fed-batch cultivations with different growth profiles and mean fluorescence per cell (open symbols) analysed by flow cytometry. Circles μ=0.1 h-1, squares μ=0.2 h-1, triangles μ=0.4 h-1.

55 Optimisation of cultivation The last step of this study was to develop an efficient process for cultivation of cells with high-level expression of proteins at the surface. Expression was favoured by cultivation in minimal medium and there was a trend that expression was higher at higher growth rates (Fig 23). We therefore reasoned that a “repeated batch strategy”, were the cells grow at the maximal specific growth rate and where new batches of glucose are added as the glucose is consumed, would be suitable for obtaining cells with high-level expression (Fig 24).

Figure 24. Cell growth, acetic acid formation and surface expression during repeated batch cultivation. (A) Cell growth measured as optical density (filled squares), dry weight (open squares), surface expression per cell measured by flow cytometry (circles, line). (B) Specific growth rate (μ, diamonds), acetic acid concentration in the medium (triangles, line) and dissolved oxygen (DOT, line).

A furhter advantage of the repeated batch strategy is seen when looking at the fluorescence peaks from the flow cytometer analysis (Fig 25). The peaks obtained during the extended batch culture are narrower than during the fed-batch cultivations, indicating that the bacterial population is more homogenous. The peak width can be considered as a quality parameter for cells expressing proteins at their surface. As discussed before, acetic acid accumulation, due to overflow metabolism, is a problem when using the batch cultivation strategy. The acetic acid accumulation in strain 0:17ΔOmpT is however much lower than in AF1000, even though the strains are both

56 K-12 derivatives and have approximately the same growth rates, and this explains why the repetitive batch strategy can be used. In the former study (I) the cell mass was 3 g l-1 when an inhibiting concentration of acetic acid, 500 mg l-1, was reached. In the present study the cell mass was 8 g l-1 when the acetic acid reached this concentration but the cells continued to grow without severe effects on the growth rate until the cell mass reached approximately 20 g l-1. The acetic acid concentration does not reach inhibiting concentrations until the oxygen limitation of the fermentor is reached. At this point, coinciding with the DOT approaching zero, a sharp increase in the acetic acid accumulation is observed. This additional acetic acid probably originates from mixed acid fermentation, which is a result of insufficient supply of oxygen to the fermentor. The final dry weight reached, by using the repetitive batch process, was approximately 30 g l-1, and the surface expression level remained high up to a cell mass of approximately 20 g l-1 (Fig 24). For a further increase in cell mass, the fed-batch technology is needed, even though this results in a less homogenous cell population.

Figure 25. Flow cytometry analysis of antibody-binding activity of surface displayed Z from two selected samples. Black (repeated batch) and grey (fed- batch, μ=0.4 h-1)

57 3 CONCLUDING REMARKS

The potential of E. coli as host for production of recombinant proteins has not yet been fully exploited. There are several examples in the literature describing efforts to overcome problems associated with heterologous protein production in this host. The development of novel and better methods for production of recombinant proteins is essential for the purpose of structural and functional studies, but also for the industrial production of biopharmaceuticals. In this thesis, the impact of the glucose uptake rate was studied with the objective to improve and optimise production of recombinant proteins in E. coli. The glucose uptake rate was shown to influence the production of the recombinant protein in all works included in this thesis. The impact of the glucose uptake rate was however different depending on the product localisation and the protein product in question.

As an alternative to traditional fed-batch technology, we here showed that mutant strains deficient in the phosphotransferase system (PTS) could be used for control of the glucose uptake rate (I). This allows for control of the specific growth rate at a genetic level rather than by physical means (e.g., by feed rate, temperature). We suggest that the mutants will be valuable tools in small-scale parallel recombinant protein production where the fed-batch technology, for practical reasons, is not applicable. The mutants were further applied for production of integral membrane proteins (III) and were able to produce membrane proteins that were not possible to produce by the corresponding wild type strain. Our results suggest that the mutants could be useful not only for production of membrane proteins but also for other “difficult-to-express” proteins, when production benefits from a lower synthesis rate and when a high cell density is needed in order to obtain a “sufficient” amount of protein.

Process development is quite often characterised by a screening stage performed under batch conditions in small parallel scale, e.g., in microtitre plates or shake flasks. Unfortunately, the level of control of the environmental conditions in such shaken systems is low and the conditions also change rapidly as a consequence of the exponential biomass-increase. In contrast, in the production scale, the environmental

58 conditions are strictly controlled, the fed-batch technology is usually applied and the production takes place under lower growth rates in comparison to batch cultures. Thus, the results obtained from small-scale cultivations do not necessarily apply in a large-scale production format, which makes the scale up process more complex. By using the PTS-mutants, screening can be performed under fed-batch-like conditions. Potentially, this will limit the amount of work for the process development team and hopefully increase the probability of selecting conditions that will work also in the large-scale production process.

That there is a need for control of the glucose uptake rate in small-scale cultivation is evidenced by the recent development of other innovative methods. One such method is EnBase®, or enzyme-based-substrate-delivery, which has been developed as a complement to the fed-batch technology (Panula-Perälä, et al., 2008; Siurkus, et al., 2010). Glucose is released into the medium by enzymatic degradation of starch by the enzyme glucoamylase and no external feed of glucose is required. The main benefit of this method is that it results in significantly higher cell densities compared to traditional shake-flasks cultures (Siurkus, et al., 2010). However, as EnBase® is a kit- based method, it is expensive to use and since the ingredients of the different components in the kit are not fully disclosed, using the EnBase® technology as a tool in the process development might be more complicated.

One problem associated with periplasmic processes is that the product tends to leak to the medium, resulting in decreased yields if the product in the medium is not used in the down stream processing. Leakage of recombinant products to the medium was clearly affected by the feed rate of glucose (II) since leakage increased with the feed rate. Although a higher feed rate resulted in an increased specific productivity, the amount of product inside the cells was constant. In order to retain the product inside the cells, the induction should preferably take place when the feed rate is low, which would not only decrease leakage but also improve the product quality (Boström, et al., 2005; Sandén, et al., 2005). In this study, we also showed that the feed rate influenced the outer membrane composition. The total amount of outer membrane proteins decreased as the feed rate increased and further reductions in outer membrane protein accumulation was seen when a recombinant product was secreted to the periplasm.

59 Although bacterial surface display has been used extensively in various research applications, there are few studies focusing on its use as a large-scale protein production tool. Here we applied the AIDA autotransporter for surface display of a recombinant protein and studied protein expression under control of the aidA promoter (IV). We showed that cultivation in minimal salt medium resulted in higher expression levels, as did higher feed rates of glucose during fed-batch cultivation. Batch cultivation with repeated additions of glucose was distinguished as the best method to improve the surface expression homogeneity in the bacterial population. In the future, it would be interesting to change the aidA promoter to an inducible and tunable promoter. By varying the inducer concentration it will then be possible to investigate if the expression level can be further increased and also potential limitations regarding Sec-translocation and the insertion process of proteins in the outer membrane can be studied. The space in the outer membrane where the transporters accumulate is limited and it will be interesting to find out (i) how many transporters that can actually be inserted in the membrane and (ii) how many transporters that can be inserted without affecting the stability of the membrane.

Extracellular production of recombinant proteins is desirable since it allows easy purification, an environment free of cell-associated proteolytic degradation and minimized impact on the host cell from toxic proteins. Some autotransporters release their passenger to the medium and may therefore be exploited for extracellular production of recombinant proteins. It would be interesting to investigate how efficient such systems would be in terms of productivity and protein quality. The number of transporters that can be inserted in the membrane is, as discussed above, limited and furthermore, the cell has to produce one transporter for each passenger, which is disadvantageous. However, for some proteins the extra cellular medium represents the only possible method for their production, and for production of these proteins, the autotransporter-based excretion strategy might be promising.

Considering what has been presented in this thesis, the glucose uptake rate is an important parameter in recombinant protein production. It does not only effect the biomass that can be reached in a certain process but influences also parameters such as protein quality, specific productivity and leakage of recombinant proteins to the

60 medium. Therefore, it is important to be able to control the glucose uptake rate also in small-scale cultivation, and as shown in this thesis, this becomes possible by using PTS-mutants.

61 4 ABBREVIATIONS

AC Adenylate cyclase Ack Acetate kinase AMP-ACS AMP-forming acetyl CoA-synthetase AIDA Adhesin involved in diffuse adherence AMP Adenosine monophosphate ATP Adenosine triphosphate BAM β-barrel assembly machinery cAMP Cyclic adenosine monophosphate CCR Carbon catabolite repression cDNA Complementary DNA CL Cardiolipin CRP cAMP receptor protein DNA Deoxyribonucleic acid DOT Dissolved oxygen tension DR Down stream region DSB Disulphide bond DW Dry weight (g l-1) E. coli Escherichia coli EI Enzyme I EII Enzyme II F Flow rate (L/h) Fab Antibody fragment HAc Acetic acid

His6 Hexahistidine tag HPr Phosphohistidine carrier protein HTP High–throughput production HTS High-throughput screening HUGO The Human Genome Organisation IgG Immunoglobulin G IMAC Immobilized metal affinity chromatography IMP Integral membrane proteins IPTG Isopropyl-β-D-thiogalactopyranoside kDa kilo Dalton LamB Maltose outer membrane porin LB Luria broth Lpp Murein lipoprotein LPS Lipopolysaccaride MBP Maltose binding protein mRNA Messenger ribonucleic acid OD Optical density OMP Outer membrane protein PE Phosphatidyletanolamine PEP Phosphoenolpyruvate PG Phosphatidylglycerol PoxB Pyruvate oxidase B ppGpp Guanosine tetraphoshate

62 PPIase Peptidyl-prolyl isomerase PrEST Protein epitope signature tag Pta Phosphotransacetylase PTS Phosphotransferase system -1 -1 qp Specific productivity (g g h ) -1 -1 qs Specific substrate consumption rate (g g h ) RNA Ribonucleic acid rRNA Ribosomal ribonucleic acid SD Shine Dalgarno sequence SDS-PAGE Sodium dodecylsulfate polyacrylamide gel electrophoresis -1 Si Substrate concentration (g l ) SRP Signal recognition particle TAT Twin arginine translocation TCA Tricarboxylic acid cycle TF Trigger factor TIR Translation initiation region tRNA Transfer ribonucleic acid V Cultivation volume (l) wt Wild type X Cell dry weight (g l-1) µ Specific growth rate (h-1) σ24 = σE Extra-cytoplasmic stress sigma factor σ32= σH Heat shock sigma factor σ38= σs Starvation/stationary phase sigma factor

63 5 ACKNOWLEDGMENT

Allting har ett slut, så även en doktorandtid, och jag skulle därför vilja ta tillfället i akt att tacka alla Er som på ett eller annat sätt stöttat mig på vägen till examen. Ett speciellt tack till:

Vinnova, Vetenskapsrådet och KTH för finansiellt stöd.

Professor Gen Larsson, för handledning och stöd genom åren, för att du alltid delar med dig av dina kunskaper och idéer samt för att du är så entusiastisk över vår forskning.

Professor Pär Nordlund och Marina Ignatushchenko och övriga medarbetare på MBB, Karolinska Institutet. Tack för att jag fick möjlighet att komma till Ert labb och genomföra min studie. Det var både lärorikt, roligt och trevligt att arbeta tillsammans med er.

Dominic Reeks, Neil Weir and Leigh Bowering at Celltech for great cooperation.

Katrin för ett ypperligt samarbete i början av doktorandtiden. Allt blir så mycket lättare och roligare när man är två.

Martin för att du fortsatte ytexpressionsprojektet när jag labbade på KI, samt för ett finfint samarbete därefter. Tack också för att du alltid ställer upp och löser mina datorrelaterade problem och för att du ritar så fina bilder. Din hjälp nu i “skrivartider” har verkligen varit värdefull.

Patrik för handledning i autotransportprojektet och för ovärderlig hjälp med avhandlingen.

Examensarbetarna Maria, Sofia och Cecilia. Det var givande att jobba tillsammans med er och tack för era bidrag till mina projekt.

Andres för hjälp med proteinrening och för att du alltid tar dig tid att diskutera forskningsrelaterade frågor. Per-Åke Nygren och Per-Åke Löfdahl för kloningstips, John Löfblom för hjälp med FACS-analys.

Alla nuvarande och före detta kollegor och vänner på avdelningen (ingen nämnd, ingen glömd) för trevligt sällskap och roliga stunder under årens lopp, samt tack för all hjälp och goda råd.

Ett extra tack till familjen som alltid hjälper mig när det behövs. Pappa och Ida för att ni lyssnar och förstår. Irene, Sören, Ida och Tobias för att ni varit barnvakt så att jag kunnat fokusera på att skriva. Jakob och Ines för att ni sprider så mycket glädje omkring er. Och sist men inte minst, ett stort tack till dig Lars för ditt tålamod, för att du är en fantastisk “hemma-pappa” och för att du alltid finns där.

64 6 REFERENCES

Abrahmsén, L., Moks, T., Nilsson, B. and Uhlén, M. (1986) Secretion of heterologous gene products to the culture medium of Escherichia coli, Nucleic Acids Research, 14, 7487-7500. Adams, J.M. (1968) On the Release of the Formyl Group from Nascent Protein, Journal of Molecular Biology, 33, 571-589. Allen, S.P., Polazzi, J.O., Gierse, J.K. and Easton, A.M. (1992) Two Novel Heat Shock Genes Encoding Proteins Produced in Response to Heterologous Protein Expression in Escherichia coli, Journal of Bacteriology, 174, 6938-6947. Arditti, R., Grodzicker, T. and Beckwith, J. (1973) Cyclic Adenosine Monophosphate-Independent Mutants of the Lactose Operon on Escherichia coli, Journal of Bacteriology, 114, 652-655. Arséne, F., Tomoyasu, T. and Bukau, B. (2000) The heat shock response of Escherichia coli, International Journal of Food Microbiology, 55, 3-9. Baars, L., Wagner, S., Wickström, D., Klepsch, M., Ytterberg, J., van Wijk, K.J. and de Gier, J.-W. (2008) Effects of SecE Depletion on the Inner and Outer Membrane Proteomes of Escherichia coli, Journal of Bacteriology, 190, 3505-3525. Bandmann, N., Collet, E., Leijen, J., Uhlén, M., Veide, A. and Nygren, P.-Å. (2000) Genetic engineering of the Fusarium solani pisi lipase cutinase for enhanced partitioning in PEG-phosphate aqueous two-phase systems, Journal of Biotechnology, 79, 161-172. Baneyx, F. (1999) Recombinant protein expression in Escherichia coli, Current Opinion in Microbiology, 10, 411-421. Baneyx, F. and Georgiou, G. (1990) In vivo degradation of secreted fusion proteins by the Escherichia coli outer membrane protease OmpT, Journal of Bacteriology, 172, 491-494. Baneyx, F. and Mujacic, M. (2004) Recombinant protein folding and misfolding in Escherichia coli, Nature Biotechnology, 22, 1399-1408. Bauer, K.A., Ben-Bassat, A., Dawson, M., De La Puente, V. and Neway, J.O. (1990) Improved Expression of Human Interleukin-2 in High-Cell-Density Fermentor Cultures of Escherichia coli K-12 by a Phosphotransacetylase Mutant, Applied and Environmental Microbiology, 56, 1296-1302. Benz, I., van Alen, T., Bolte, J., Wörmann, M.E. and Schmidt, A.M. (2010) Modulation of transcription and characterization of the promoter organization of the autotransporter adhesin heptosyltransferase and the autotransporter adhesin AIDA-I, Microbiology, 156, 1155-1166. Berks, B.C., Palmer, T. and Sargent, F. (2005) Protein targeting by the bacterial twin-arginine translocation (Tat) pathway, Current Opinion in Microbiology, 8, 174-181. Blight, M.A. and Holland, B. (1994) Heterologous protein secretion and the versatile Escherichia coli haemolysin translocator, Tibtech, 12, 450-455. Booth, I.R. (1999) Adaptation to Extreme Environments, In: Biology of the Prokaryotes, Eds: J W Lengeler, G Drews, H G Schlegel (Blackwell Science), 652-671. Boström, M., Markland, K., Sandén, A.M., Hedhammar, M., Hober, S. and Larsson, G. (2005) Effect of substrate feed rate on recombinant protein secretion, degradation and inclusion body formation in Escherichia coli, Applied Microbiology and Biotechnology, 68, 82-90. Casadaban, M.J. (1976) Transposition and Fusion of the lac Genes to Selected Promoters in Escherichia coli using Bacteriophage Lambda and Mu, Journal of Molecular Biology, 104, 541-555. Charbit, A., Molla, A., Saurin, W. and Hofnung, M. (1988) Versatility of a vector for expressing foreign polypeptides at the surface of Gram-negative bacteria, Gene, 70, 181-189. Chaudhiri, T.K., Horii, K., Yoda, T., Arai, M., Nagata, S., Terada, T., Uchiyama, H., Ikura, T., Tsumoto, K., Kataoka, H., Matsushima, M., Kuwajima, K. and Kumagai, I. (1999) Effect of the Extra N-terminal Methionine Residue on the Stability and Folding of Recombinant alpha-lactalbumin Expressed in Escherichia coli, Journal of Molecular Biology, 285, 1179-1194.

65 Chen, R. and Henning, U. (1996) A periplasmic protein (Skp) of Escherichia coli selectively binds a class of outer membrane proteins, Molecular Microbiology, 19, 1287-1294. Chen, Y., Song, J., Sui, S.-f. and Wang, D.-N. (2003) DnaK and DnaJ facilitated the folding process and reduced inclusion body formation of magnesium transporter CorA overexpressed in Escherichia coli, Protein Expression and Purification, 32, 221-231. Cho, S., Shin, D., Ji, G.E., Heu, S. and Ryu, S. (2005) High-level recombinant protein production by overexpression of Mlc in Escherichia coli, Journal of Biotechnology, 119, 197-203. Chou, C.-H., Bennett, G.N. and San, K.-Y. (1994) Effect of Modified Glucose Uptake Using Genetic Engineering Techniques on High-Level Recombinant Protein Production in Escherichia coli Dense Cultures, Biotechnology and Bioengineering, 44, 952-960. Cohen, S., N, Chang, A.C.Y., Boyer, H.W. and Helling, R.B. (1973) Construction of Biologically Functional Bacterial Plasmids In Vitro, Proceedings of the National Academy of Sciences of the United States of America, 70, 3240-3244. Corchero, J.L., Viaplana, E., Benito, A. and Villaverde, A. (1996) The position of the heterologous domain can influence the solubiity and proteolysis of β-galactosidase fusion proteins in E. coli, Journal of Biotechnology, 48, 191-200. Cronan, J.E., JR and O.Rock, C. (1996) Biosynthesis of Membrane Lipids, In Escherichia coli and Salmonella, Cellular and Molecular Biology, Ed: Neidhart F C (ASM Press), 612-636. Daugherty, P.S. (2007) Protein engineering with bacterial display, Current Opinion in Structural Biology, 17, 474-480. Dautin, N. and Bernstein, H.D. (2007) Protein Secretion in Gram-Negative Bacteria via the Autotransporter Pathway, The Annual Review of Microbiology, 61, 89-112. De Anda, R., Lara, A.R., Hernández, V., Hernández-Montalvo, V., Gosset, G., Bolivar, F. and Ramirez, O.T. (2006) Replacement of the glucose phospotransferase system by galactose permease reduces acetate accumulation and improves process performance of Escherichia coli for recombinant protein production without impairment of growth rate, Metabolic Engineering, 8, 281-290. Death, A., Notley, L. and Ferenci, T. (1993) Derepression of LamB Protein Facilitates Outer membrane Permeation of Carbohydrates into Escherichia coli under Conditions of Nutrient Stress, Journal of Bacteriology, 1475-1483. Dong, H., Nilsson, L. and Kurland, C.G. (1995) Gratuitous Overexpression of Genes in Escherichia coli Leads to Growth Inhibition and Ribosome Destruction, Journal of Bacteriology, 177, 1497-1504. Dowhan, W. (1997) Molecular basis for membrane phospholipid diversity: Why Are There So Many Lipids?, Annual Review of Biochemistry, 66, 199-232. du Plessis, D.J.F., Nouwen, N. and Driessen, A.J.M. (2011) The Sec translocase, Biochimica et Biophyisca Acta, 1808, 851-865. Eiterman, M.A. and Altman, E. (2006) Overcoming acetate in Escherichia coli recombinant protein fermentations, TRENDS in Biotechnology, 24, 530-536. Enfors, S.-O. and Häggström, L. (2000) Bioprocess technology fundamentals and applications Royal Institute of Technology (KTH), Stockholm. Eshagi, S., Hedrén, M., Ignatushenko Abdel Nasser, M., Hammarberg, T., Thornell, A. and Nordlund, P. (2005) An efficient strategy for high-throughput expression screening of recombinant integral membrane proteins, Protein Science, 14, 676-683. Essen, L.-O. (2002) Structural genomics of "non-standard" proteins: a chance for membrane proteins?, Gene Function & Disease, 3, 39-48. Francisco, J.A., Campbell, R., Iverson, B.L. and Georgiou, G. (1993) Production and flourescence- activated cell sorting of Escherichia coli expressing a functional antibody fragment on the external surface, Proceedings of the National Academy of Sciences of the United States of America, Biochemistry, 90, 10444-10448. Genentech (press release, 1978) Available from:, http://www.gene.com/gene/news/press- releases/display.do?method=detail&id=4160, 2011-05-10.

66 Georgiou, G. and Shuler, M.L. (1988) Release of Periplasmic Enzymes and Other Physiological Effects of ß-lactamase Overproduction in Escherichia coli, Biotechnology and Bioengineering, 32, 741-748. Golomb, M. and Chamberlin, M. (1974) Characterization of T7-specific Ribonucleic Acid Polymerase. IV. Resolution of the major in vitro transcripts by gel electrophoresis., The Journal of Biological Chemistry, 249, 2858-2863. Gosset, G. (2005) Improvement of Escherichia coli production strains by modification of the phosphoenolpyruvate:sugar phosphotransferase system, Microbial Cell Factories, 4:14. Gross, C.A. (1996) Function and Regulation of the Heat Shock Proteins, In: Escherichia coli and Salmonella, Cellular and molecular biology, Eds: Neidhart F C (ASM Press). Hannig, G. and Makrides, S.C. (1998) Strategies for optimizing heterologous protein expression in Escherichia coli, Tibtech, 9, 54-60. Hauke, L., Schwarz, E. and Rudolph, R. (1998) Advances in refolding of proteins produced in E. coli, Current Opinion in Biotechnology, 9, 497-501. Hennecke, G., Nolte, J., Volkmer-Engert, R., Schneider-Mergener, J. and Behrens, S. (2005) The periplasmic chaperone SurA exploits two features characteristic of integral outer membrane proteins for selective substrate recognition, The Journal of Biological Chemistry, 280, 23540-23548. Hoffmann, F. and Rinas, U. (2004) Stress induced by recombinant protein production in Escherichia coli, In: Physiological Stress Responses in Bioprocesses, Advances in Biochemical Engineering/Biotechnology, 89, Ed Enfors S. O (Springer). Jensen, E.B. and Carlsen, S. (1990) Production of Recombinant Human Growth Hormone in Escherichia coli: Expression of Different Precursors and Physiological Effects of Glucose, Acetate and Salts, Biotechnology and Bioengineering, 36, 1-11. Jose, J. (2006) Autodisplay: efficient bacterial surface display of recombinant proteins, Applied Microbiology and Biotechnology, 69, 607-614. Jose, J., Bernhardt, R. and Hannemann, F. (2002) Cellular surface display of dimeric Adx and whole cell P450-mediated steroid synthesis on E. coli Journal of Biotechnology, 95, 257-268. Jose, J., Chung, J.-W., Jeon, B.-J., Maas, R.M., Nam, C.-H. and Pyun, J.-C. (2009) Escherichia coli with autodisplayed Z-domain of protein A for signal amplification of SPR biosensor, Biosensors and Bioelectronics, 24, 1324-1329. Jose, J. and Meyer, T.F. (2007) The Autodisplay Story, from Discovery to Biotechnical and Biomedical Applications, Microbiology and Molecular Biology Reviews, 71, 600-619. Jose, J. and von Schwichow, S. (2004) Autodisplay of Active Sorbitol Dehydrogenase (SDH) Yields a Whole Cell Biocatalyst for the Synthesis of Rare Sugars, ChemBioChem, 5, 491-499. Jürgen, B., Breitenstein, A., Urlacher, V., Büttner, K., Hongying, L., Hecker, M., Schweder, T. and Neubauer, P. (2010) Quality control of Inclusions bodies in Escherichia coli, Microbial Cell Factories, 9:41. Kiefer, H. (2003) In vitro folding of alpha-helical membrane proteins, Biochimica et Biophyisca Acta, 1610, 57-61. Klauser, T., Pohlner, J. and Meyer, T.F. (1990) Extracellular transport of cholera toxin B subunit using Neisseria IgA protease β-domain: conformation-dependent outer membrane translocation, The EMBO Journal, 9, 1991-1999. Klemm, P. and Schembri, M., A (2000) Fimbrial surface display systems in bacteria: from vaccines to random libraries, Microbiology, 146, 3025-3032. Knowles, T.J., Scott-Tucker, A., Overduin, M. and Henderson, I.R. (2009) Membrane protein architects: the role of the BAM complex in outer membrane protein assembly, Nature Reviews Microbiology, 7, 206-214. Koh, B.T., Nakashimada, U., Pfeiffer, M. and Yap, M.G.S. (1992) Comparision of acetate inhibition on growth of host and recombinant E. coli K12 strains, Biotechnology Letters, 14, 1115-1118.

67 Kramer, R.A., Zandwijken, D., Egmond, M.R. and Dekker, N. (2000) In vitro folding, purification and characterization of Escherichia coli outer membrane protease OmpT, European Journal of Biochemistry, 267, 885-893. Lara, A.R., Caspeta, L., Gosset, G., Bolivar, F. and Ramirez, O.T. (2008) Utility of an Escherichia coli Strain Engineered in the Substrate Uptake System for Improved Culture Performance at High Glucose and Cell Concentrations: An alternative to Fed-Batch Cultures, Biotechnology and Bioengineering, 99, 893-901. Lazzaroni, J.-C. and Portalier, R.C. (1981) Genetic and Biochemical Characterization of periplasmic- Leaky Mutants of Escherichia coli K-12, Journal of Bacteriology, 145, 1351-1358. Lee, S.Y. (1996) High cell-density culture of Escherichia coli, Tibtech, 14, 98-105. Lee, S.Y., Choi, J.H. and Xu, Z. (2003) Microbial cell-surface display TRENDS in Biotechnology, 21, 45-52. Lengeler, J.W. and Postma, P.W. (1999) Global regulatory networks and signal transduction pathways, In: Biology of the Prokaryotes, Eds: J W Lengeler, G Drews, H Schlegel (Blackwell Science) 491-523. Li, C., Zhu, Y., Benz, I., Schmidt, A., Chen, W., Mulchandani, A. and Qiao, C. (2008) Presentation of functional organophosphorus hydrolase fusions on the surface of Escherichia coli by the AIDA-I autotransporter pathway, Biotechnology and Bioengineering, 99, 485-490. Lobban, P.E. and Kaiser, A.D. (1973) Enzymatic end-to end joining of DNA molecules, Journal of Molecular Biology, 78, 453-471. Löfdahl, S., Guss, B., Uhlén, M., Philipson, L. and Lindberg, M. (1983) Gene for staphylococcal protein A, Proceedings of the National Academy of Sciences of the United States of America, Biochemistry, 80, 697-701. Luckey, M. (2008), Membrane Structural Biology, chapter 1 (Cambridge University Press). Magnusson, L.U., Farewell, A. and Thomas, N. (2005) ppGpp: a global regulator in Escherichia coli, TRENDS in Microbiology, 13, 236-242. Makrides, S.C. (1996) Strategies for Achieving High-level Expression of genes in Escherichia coli, Micobiological Reviews, 60, 512-538. Mannesse, M.L.M., Cox, R.C., Koops, B.C., Verheij, H.M., de Haas, G.H., Egmond, M.R., van der Hijden, H.T.W.M. and de Vlieg, J. (1995) Cutinase from Fusarium solani pisi Hydrolyzing Triglyceride Analogues. Effect of Acyl Chain Length and Position in the Subtrate Molecule on Activity and Enantioselectivity, Biochemistry, 34, 6400-6407. Martínez-Alonso, M., García-Fruitós, E., Ferrer-Miralles, N., Rinas, U. and Villaverde, A. (2010) Side effects of chaperone gene co-expression in recombinant protein production, Microbial Cell Factories, 9:64. Maurer, J., Jose, J. and Meyer, T.F. (1997) Autodisplay: One-component system for efficient surface display and release of soluble recombinant proteins from Escherichia coli, Journal of Bacteriology, 179, 794-804. Mergulhão, F.J.M. and Monteiro, G.A. (2004) Secretion Capacity Limitations of the Sec Pathway in Escherichia coli, Journal of Microbiology and Biotechnology, 14, 128-133. Mergulhão, F.J.M., Summers, D.K. and Monteiro, G.A. (2005) Recombinant protein secretion in Escherichia coli, Biotechnology Advances, 23, 177-202. Meyer, H.-P., Leist, C. and Fiechter, A. (1984) Acetate formation in continuous culture of Escherichia coli K12 D1 on defined and complex media, Journal of Biotechnology, 1, 355-358. Miksch, G., Neitzel, R., Fiedler, E., Friehs, K. and Flaschel, E. (1997) Extracellular production of a hybrid β-glucanase from Bacillus by Escherichia coli under different cultivation conditions in shaking cultures and bioreactors, Applied and Environmental Microbiology, 47, 120-126. Miroux, B. and Walker, J.E. (1996) Over-production of Proteins in Escherichia coli: Mutant Hosts that Allow Synthesis of some Membrane Proteins and Globular Proteins at High Levels, Journal of Molecular Biology, 260, 289-298.

68 Missiakas, D., Betton, J.-M. and Raina, S. (1996) New components of protein folding in extracytoplasmic compartments of Escherichia coli SurA, FkpA and Skp/OmpH, Molecular Microbiology, 21, 871-884. Moat, A.G., Foster, J.W. and Spector, M.P. (2002), Microbial Physiology, (Wiley-Liss) Moerschell, R.P., Hosokawa, Y., Tsunasawa, S. and Sherman, F. (1990) The Specificities of Yeast Methionine Aminopeptidase and Acetylation of Amino-terminal Methionine in Vivo, The Journal of Biological Chemistry, 265, 19638-19643. Morrow, J.F., Cohen, S., N, Chang, A.C.Y., Boyer, H.W., Goodman, H.M. and Helling, R.B. (1974) Preplication and Transcription of Eukaryotic DNA in Escherichia coli, Proceedings of the National Academy of Sciences of the United States of America, 71, 1743-1747. Nakano, K., Rischke, M., Sato, S. and Märkl, H. (1997) Influence of acetic acid on the growth of Escherichia coli K12 during high-cell-density cultivation in a dialysis reactor, Applied Microbiology and Biotechnology, 48, 597-601. Neidhardt, F.C., Ingraham, J.L. and Schaechter, M. (1990), Physiology of the Bacterial Cell, A Molecular Approach, chapter 1 (Sinauer Associates). Ni, Y. and Chen, R. (2009) Extracellular recombinant protein production from Escherichia coli, Biotechnology Letters, 31, 1661-1670. Nikaido, H. (1996) Outer Membrane, In: Escherichia coli and Salmonella, Cellular and molecular biology, Eds: Neidhart F C (ASM Press), 29-47. Nikaido, H., Bavoil, P. and Hiroa, Y. (1977) Outer membranes of gram-negative bacteria. XV. Transmembrane diffusion rates in lipoprotein-deficient mutants of Escherichia coli, Journal of Bacteriology, 132, 1045-1047. Olsson, M.O. and Isaksson, L.A. (1979) Analysis of rpsD Mutations in Escherichia coli. I: Comparison of Mutants with Various Alterations in Ribosomal Protein S4, Molecular & General Genetics, 1, 251- 257. Panula-Perälä, J., Siurkus, J., Vasala, A., Wilmanowski, R., Casteleijn, M.G. and Neubauer, P. (2008) Enzyme controlled glucose auto-delivery for high cell density cultivations in microplates and shake flasks, Microbial Cell Factories, 7:31. Phue, J.-N., Lee, S.J., Kaufman, B., Alejandro, N. and Shiloach, J. (2010) Acetate accumulation through alternative metabolic pathways in ackA- pta- poxB - triple mutant in E. coli B (BL21), Biotechnology Letters, 32, 1897-1903. Phue, J.-N. and Shiloach, J. (2004) Transcription levels of key metabolic genes are the cause fo the different glucose utilization pathways in E. coli B (BL21) and E. coli K (JM109), Journal of Biotechnology, 109, 21-30. Picon, A., Texiera de Mattos, M.J. and Postma, P.W. (2005) Reducing the Glucose Uptake Rate in Escherichia coli Affects Growth Rate But Not Protein Production, Biotechnology and Bioengineering, 90, 191-200. Plumbridge, J. (2002) Regulation of gene expression in the PTS in Escherichia coli: the role and interactions of Mlc, Current Opinion in Microbiology, 5, 187-193. Postma, P., Lengeler, J. and Jacobsson, G. (1996) Phosphoenolpyruvate: carbohydrate phosphotransferase systems (PTS), In: Escherichia coli and Salmonella typhimurium. Cellular and Molecular Biology, Ed: Neidhardt F. C (ASM Press). Ryan, W., Collier, P., Loredo, L., Pope, J. and Sachdev, R. (1996) Growth Kinetics of Escherichia coli and Expression of a recombinant Protein and Its Isoforms under Heat Shock Conditions, Biotechnology Progress, 12, 596-601. Sandén, A.M., Boström, M., Markland, K. and Larsson, G. (2005) Solubility and Proteolysis of the Zb- MalE and Zb-MalE31 Proteins During Overproduction in Escherichia coli, Biotechnology and Bioengineering, 90, 239-247. Sandén, A.M., Prytz, I., Tubulekas, I., Förberg, C., Le, H., Hektor, A., Neubauer, P., Pragai, Z., Harwood, C., Ward, A., Picon, A., Texiera de Mattos, J., Postma, P., Farewell, A., Nyström, T., Reeh,

69 S., Pedersen, S. and Larsson, G. (2002) Limiting Factors in Escherichia coli Fed-Batch Production of Recombinant Proteins, Biotechnology and Bioengineering, 81, 158-166. Schmidt, F.R. (2004) Recombinant expression systems in the pharmaceutical industry, Applied Microbiology and Biotechnology, 65, 363-372. Shiloach, J. and Fass, R. (2005) Growing E. coli to high cell density - A historical perspective on method development, Biotechnology Advances, 23, 345-357. Shokri, A., Sandén, A.-M. and Larsson, G. (2002) Growth rate-dependent changes in Escherichia coli membrane structure and protein leakage, Applied Microbiology and Biotechnology, 58, 386-392. Shokri, A., Sandén, A.-M. and Larsson, G. (2003) Cell and process design for targeting of recombinant proteins into the culture medium of Escherichia coli, Applied Microbiology and Biotechnology, 60, 654-664. Singer, S.J. and Nicolson, G.L. (1972) The Fluid Mosaic Model of the Structure of Cell Membranes, Science, 175, 720-731. Siurkus, J., Panula-Perälä, J., Horn, U., Kraft, M., Rimseliene, R. and Neubauer, P. (2010) Novel approach of high cell density recombinant bioprocess development: Optimasation and scale-up from microtitre to pilot scales while maintaining the fed-batch cultivation mode of E. coli cultures, Microbial Cell Factories, 9:35. Sklar, J.G., Wu, T., Kahne, D. and Silhavy, T.J. (2007) Defining the roles of the periplasmic chaperones SurA, Skp and DegP in Escherichia coli, Genes and Development, 21, 2473-2484. Sonoda, H., Kumada, Y., Katsuda, T. and Yamaji, H. (2011) Effects of cytoplasmic and periplasmic chaperones on secretory production of single-chain Fv antibody in Escherichia coli, Journal of Bioscience and Bioengineering, 111, 465-470. Sorensen, H.P. and Mortensen, K.K. (2005) Soluble expression of recombinant proteins in the cytoplasm of Escherichia coli, Microbial Cell Factories, 4:1. Sørensen, H.P. and Mortensen, K.K. (2005) Advanced genetic strategies for recombinant protein expression in Escherichia coli, Journal of Biotechnology, 115, 113-128. Stathopoulos, C., Georgiou, G. and Earhart, C., F (1996) Characterization of Escherichia coli expressing an Lpp´OmpA(46-159)-PhoA fusion protein localized in the outer membrane, Applied Microbiology and Biotechnology, 45, 112-119. Stenström, M.C. and Isaksson, L.A. (2002) Influences on translation initiation and early elongation by the messenger RNA region flanking the initiation codon at the 3´side, Gene, 288, 1-8. Studier, F.W. and Moffatt, B.A. (1986) Use of Bacteriophage T7 RNA Polymerase to Direct Selective High-level Expression of Cloned Genes, Journal of Molecular Biology, 189, 113-130. Swartz, J.R. (2001) Advances in Escherichia coli production of therapeutic proteins, Current Opinion in Biotechnology, 12, 195-201. Tamaki, S., Sato, T. and Matsuhashi, M. (1971) Role of lipopolysaccharides in antibiotic resistance bacteriophage adsorption of Escherichia coli K-12, Journal of Bacteriology, 968-975. Tegel, H., Ottosson, J. and Hober, S. (2011) Enhancing the protein production levels in Escherichia coli with a strong promoter, FEBS Journal, 278, 729-739. Terpe, K. (2006) Overview of bacterial expression systems for heterologous protein production: from molecular and biochemical fundamentals to commercial systems, Applied Microbiology and Biotechnology, 72, 211-222. Turner, C., Gregory, M.E. and Turner, M.K. (1994) A study of the effect of specific growth rate and acetate on recombinant protein production of Escherichia coli JM107, Biotechnology Letters, 16, 891- 896. van Bloois, E., Winter, R.T., Kolmar, H. and Fraaije, M.W. (2011) Decorating microbes: surface display of proteins on Escherichia coli, TRENDS in Biotechnology, 29, 79-86. van de Walle, M. and Shiloach, J. (1998) Proposed Mechanism of Acetate Accumulation in Two Recombinant Escherichia coli Strains During High Density Fermentation, Biotechnology and Bioengineering, 57, 71-78.

70 Wagner, S., Baars, L., Ytterberg, J., Klussmeier, A., Wagner, C.S., Nord, O., Nygren, P.Å., van Wijk, K.J. and de Gier, J.-W. (2007) Consequences of Membrane Protein Overexpression in Escherichia coli, Molecular & Cellular Proteomics, 1527-1549. Wagner, S., Klepsch, M.M., Schlegel, S., Appel, A., Draheim, R., Tarry, M., Högbom, M., van Wijk, K.J., Slotboom, D.J., Persson, J.O. and de Gier, J.-W. (2008) Tuning Escherichia coli for membrane protein overexpression, Proceedings of the National Academy of Sciences of the United States of America, 105, 14371-14376. Wagner, S., Lerch Bader, M.L., Drew, D. and De Gier, J.W. (2006) Rationalizing membrane protein overexpression, TRENDS in Biotechnology, 24, 364-371. Walton, T.A., Sandoval, C.M., Fowler, C.A., Pardi, A. and Sousa, M.C. (2009) The cavity-chaperone Skp protects its substrate from aggregation but allows independent folding of substrate domains, Proceedings of the National Academy of Sciences of the United States of America, 106, 1771-1777. Weibezahn, J., Bukau, B. and Mogk, A. (2004) Unscrambling an egg: protein disaggregation by AAA+ proteins, Microbial Cell Factories, 3:1. White, C.B., Chen, Q., Kenyon, G.L. and Babbit, P.C. (1995) A novel activity of OmpT. Proteolysis under extreme denaturing conditions, The Journal of Biological Chemistry, 270, 12990-12994. Wilson Bowers, C., Lau, F. and Silhavy, T.J. (2003) Secretion of LamB-LacZ by the Signal Recognition Particle Pathway of Escherichia coli, Journal of Bacteriology, 185, 5697-5705. Wolfe, A.J. (2005) The acetate Switch, Microbiology and Molecular Biology Reviews, 69, 12-50. Young, J.C., Agashe, V.R., Siegers, K. and Hartl, F.U. (2004) Pathways of chaperone-mediated protein folding in the cytosol, Nature Reviews Molecular Cell Biology, 5, 781-791.